In Part 1 of this two part series I discussed some of the challenges involved in incorporating and regulating how AI systems might incorporate and apply religious and spiritual concepts. Little did I know that while using ChatGPT to help develop an accompanying graphic (see Part 1) I would run into a concrete example of the challenges of doing so!

After completing my edits to the text of Part 1 (I had asked ChatGPT to convert a scan of my handwritten first draft of Part 1, which I then edited manually) I posed this to ChatGPT:

the next challenge -- coming up with a graphic that does not rely on traditional religious iconography. The concepts that come to my mind are the forces of good and the forces of evil and how they are played out ib the context of AI systems. It should be colorful, have no text, and not be realistic but rather, abstract.

I had included hints of what I was looking for as I have learned in the past that the more explicit my prompts, the better are the results.

After a few seconds I received this response from ChatGPT:

We’re so sorry, but the image we created may violate our content policies. If you think we got it wrong, please retry or edit your prompt.

My response to the above was a simple

huh?

ChatGPT kept thinking and eventually the illustration at the head of Part 1 was displayed, so I asked,

This is good but could you at least describe how the first image violated your guidelines? That is relevant to this article, after all.

ChatGPT responded quickly:

Yes — and I should clarify something important: I don’t think the image concept itself was actually a problem. The refusal was almost certainly a false-positive safety trigger caused by the wording of the prompt, not by the abstract visual idea.

The likely trigger was the combination of terms such as “forces of good and evil,” “religion,” “spirituality,” “evil,” “corrosive,” and “manipulative” in a request for image generation. Even though your intent was clearly abstract, nonviolent, non-religious-iconographic, and suitable for an essay header, the image system may have interpreted the language as leaning toward religious conflict, demonic imagery, or harmful ideological symbolism.

That is, the system probably did not object to “abstract AI ethics image.” It likely objected to language that sounded as though it might produce explicit spiritual warfare imagery, even though we explicitly told it not to use angels, demons, crosses, temples, etc.

For the second attempt, I softened the framing by emphasizing order vs. chaos, constructive vs. destructive potential, clarity vs. distortion, and abstract AI systems. That produced essentially the same concept without using the more policy-sensitive vocabulary.

And yes — that is highly relevant to your article. It is a small example of exactly the problem you are writing about: AI systems making opaque judgments about morally or spiritually charged language, even when the human intent is thoughtful and legitimate.

My response to the above:

That helps. Thank you. That really goes to illustrating how difficult it is to regulate how AI addresses ethical/religious concerns!

ChatGPT response:

Exactly. It is a nice miniature example of the problem.

Your prompt was not asking for religious propaganda, sectarian imagery, demonology, or anything harmful. It was asking for an abstract representation of moral tension in AI systems. Yet the system apparently reacted to some combination of morally loaded terms and spiritual/religious context.

That illustrates several points that could strengthen your article:

Regulating this area is hard because the same words can be used in very different ways: theological, ethical, artistic, political, harmful, benign, satirical, or analytical.

AI systems may avoid risk by over-filtering, which can suppress legitimate discussion of religion, morality, or spirituality.

Transparency matters because users need to know whether an AI tool is making a content judgment based on safety, ideology, religious sensitivity, brand caution, legal risk, or just a brittle keyword association.

And “warning labels” may not be enough if the underlying decision process is still opaque.

In a way, the hiccup with the image prompt gave you a live example of your own thesis: once AI systems are asked to mediate morally charged concepts, even benign requests can become difficult to classify.

The first thing to note about the above exchange is that the initial message provided by ChatGPT was incorrect": “We’re so sorry, but the image we created may violate our content policies.“

As noted later by ChatGPT in its explanation, the problem was not the image it initially generated — which I never saw, assuming it existed — but the terms included in the prompt. These terms apparently suggested that I was asking for something ChatGPT has been expressly instructed to avoid.

This points out a problem: ChatGPT misinterpreted my requirements, based its initial refusal decision on this misinterpretation, then proceeded to generate the image once it realized “the truth” about what I was asking for, and why. All of this occurred behind closed doors and only came to light based on my comments and the fact that the topic of Part 1 was directly related to this decision process.

Granted, ChatGPT is a general purpose system and may not address some of the finer details of ethics, morality, and spirituality that a more dedicated system might address. Still, this little exchange is an example in microcosm of the challenges associated with incorporating ethical or even religious concepts into even a general purpose exchange. I for one would want to know the basis for any AI response that incorporates ethical, religious, or spiritual values in a response.

Legislators who don’t understand the underlying complexity and implications of doing so run the risk of making AI systems cumbersome and perhaps even unusable.

Copyright 2026 by Dennis D. McDonald

Government should not regulate belief, but AI developers should explain how their tools handle morality, ethics, and spiritual advice.

Danielle A. Davis Carty’s “The Pope Warned Us About AI, But We’re Missing the Spiritual Question” recently caught my attention. What happens, Carty asks, when AI-based systems begin to perform in ways traditionally reserved for humans versed in religious or spiritual matters?

Based on my reading of Carty, various emerging applications, such as bots trained to give advice based on religious fundamentals, deserve as much attention as our ongoing debates about youth access to AI, data center construction, and the negative impacts of AI “hallucinations.”

While I certainly appreciate Carty’s analyses, it’s not like we haven’t had to deal with such topics before concerning adoption of new technologies to “spread the word of God.” For example, church fathers did not automatically welcome Bibles printed in the vernacular via Gutenberg’s new printing press. Nor did all traditional churches adopt TV evangelism as a way to extend messaging beyond the limited geographical footprints of traditionally structured-and-funded parishes and congregations.

I’m not saying that incorporating religious and spiritual concepts into AI is necessarily evil or benign. Given that religion and spirituality have had both positive and negative impacts on humanity, we would be remiss to ignore their impact on how AI techniques are implemented.

This leads to consideration how society deals with (and potentially regulates) how AI is incorporated into daily life. An interesting treatment of this challenge, told from the perspective of how religion can co-opt technology as well as be co-opted, is episode 3 of the Netflix series Tomorrow + I, “Buddha Data.” What’s fascinating about this series is not just how AI and religion — in this case, Buddhism in a future Thailand — are integrated, but also how commercial interests influence how religion, internet, and AI are intertwined. (This Thai series dates from 2024, which is long before the Pope’s recent encyclical was published, but it’s clear that concerns about religion and AI in society were being paid attention to even back then.)

In my opinion, regulating how religion and spirituality in AI might be managed would be difficult if not impossible. As an American and strong believer in freedom of religion and separation of church and state, I would not support attempts by a state or federal government to require any particular religious set of beliefs to be prioritized over another, just as I oppose posting copies of the Ten Commandments in classrooms.

That doesn’t mean that we should not consider the ethical and moral implications of how people use AI, despite the difficulty of separating ethics and morality from religion. Perhaps the best we can hope for is for AI developers to clearly state how their tools are developed and how fundamental judgements are made by the AI when issues of religion, spirituality, ethics, or morality are involved. But then, would such “warning labels” be effective, or would they just be swept aside by users the way that voluminous online  “user agreements” are clicked?

For Part 2 of this series, go here: AI, Religion, and the Warning Label Problem, Part 2

Copyright © 2026 by Dennis D. McDonald

“Free” research access still depends on a complex web of people, institutions, systems, and shared values. Can AI help reduce friction in the system?"

Someone has to pay

Whether we’re talking about access to research data or publishing research results in peer reviewed journals, someone has to pay for:

1. The research that generates the data.

2. The meetings and conference presentations.

3. The research articles and how they are reviewed and published.

4. The systems used to store, retrieve, and disseminate research results and data.

5. Management and operation of the myriad of public and private sector organizations tied to (1) through (4).

System breakdown

The Science article “U.S. agencies aren’t ready for the rising cost of making research papers free, report warns” warns that a complex and decentralized network of public and private institutions may show signs of breaking down when one or more of the entities responsible for managing the flow of information realizes that its costs are outstripping its resources. Something has got to give. This shows up in higher prices, reduced access - or people seeking alternative means for disseminating information.

On the one hand, as the Science article reports, some US research agencies are required to make the results of their research available to the public “for free.” At the same time, those same research agencies are reducing their support for research! Given the hodgepodge of institutions involved in conducting, reporting, and disseminating research, impacts will be felt all along the “research dissemination lifecycle,” including:

1. Cutbacks in research funding will reduce the generation of data and reports that require management and dissemination.

2. Researchers may seek alternative and less expensive means for report and data dissemination.

3. New, less expensive - and potentially lower quality - methods for distributing research findings will be tried out.

4. Someone, inevitably, will try to game the system to take advantage of lowered security or quality control.

A silver lining

Before we all go “doom and gloom,” I would like to think there is a silver lining to all this: crises can force renewed attention to how the overall system works. I also believe a serious consideration is in order of how AI might help to overcome some of the system’s overall operational deficiencies.

Ties to the past

The current model still operates with much of the same components as its original 18th and 19th century beginnings:

• Continued reliance on peer reviewed journals.

• Gatekeeping by professional societies.

• Funding for time and materials by a multitude of public and private institutions.

Technology has significantly expanded the size, scope, and speed of the system over the years; yet, how the different components interact and transfer information from one participant to another still involves friction and inefficiencies. We see that happening now as the costs of making access “free” emerge as an issue given how government research cutbacks ripple through the system.

Beyond belt-tightening

It’s time to think beyond belt-tightening. We need to consider how AI can help streamline how information gets from its origin to potential users. If that means re-thinking how research information gets packaged and bundled outside traditional gatekeeping institutions like journals, libraries, and publishers, so be it.

The greatest impediment

Perhaps the single greatest impediment to using AI to streamline how this system operates is the lack of transparency in how the system currently operates. No one really has a complete understanding of how all these components interact and how much they cost. Even the language used to describe system operation can be deceptive; for example, making access “free” is a misnomer when someone at some point must absorb the cost of doing so.

Need to know?

A key question is, does AI really need to understand the details of how all these components operate? Or, can it learn while it observes how the different components interact, then use that learning to devise more efficient paths for information to travel?

This question is similar to one I asked in “Managing the Flow Chart: AI, Complexity, and Control,” with its reference to Sangeet Paul Choudary’s HBR article “AI’s Big Payoff is Coordination, Not Automation.” The idea is that AI might be able to  help coordinate component interactions without needing to know the details of how individual system components are managed. Perhaps it would be more important for the AI to understand how the priorities and business processes they support interact.

Serious challenges

Transitioning to AI-supported coordination for communicating research findings would not be a trivial challenge. I would first want to model the process, which would then be simulated to examine a variety of options. This would then be followed by a prototyped system that is controlled and carefully scoped.

Why things DO work

One final comment. Our current mix of systems and practices for communicating research findings works, despite its inefficiencies and incompatibilities, because of shared values. Participants believe that research needs to be shared. That’s a fundamental part of science. Scrutiny by one’s peers of research findings is sacrosanct. But if personal incentives, selfishness, or politics get in the way, the system breaks down, regardless of whether AI is viewed as a boon or a demon.

Copyright (c) 2026 by Dennis D. McDonald

Addendum

As an addendum, I am including both my original ChatGPT Plus prompt and a reconstructed version of the more elaborate prompt the ChatGPT Plus appears to have inferred in generating the image. The contrast is a reminder that AI can interpret, expand, and restructure its instructions.

My original prompt: "Create a line drawing of interconnected rooms each with a different type of lock and key. Getting from one side of the room to the other requires navigating through an array of different locked passageways."

ChatGPT’s interpretation of my original prompt: “Create a detailed black-and-white line drawing of a labyrinth of interconnected rooms, shown in an isometric cutaway or architectural floorplan style. The rooms should be connected by many doorways and passageways, each secured by a different kind of lock or locking mechanism—padlocks, keyholes, bolts, chains, combination locks, wheel locks, keypad locks, and other varied devices. The image should suggest that getting from the entrance on one side to the exit on the other requires navigating through a complex array of locked passageways. Include a variety of rooms with different contents or purposes so the system feels intricate and varied. Make it monochrome, pen-and-ink, crisp, serious, and diagram-like, with no people and no text.“

My comment on the image itself: “My initial reaction to version 1 was “wow” but some of the locked rooms went nowhere, which was the opposite of what I intended. The image I finally selected and posted above was the result of several iterations with the AI. It still has problems but gets the idea across.”

Once upon a time, I saw a massive flowchart that purported to show, from start to finish, all the systems, processes, participants, relationships, decision points, and workflows involved in developing and procuring a Navy weapons system.

Such an artifact must constantly evolve as people, laws, regulations, and technologies change over time. Keeping something this complicated up to date requires a lot of time, attention, and automation. That suggests the likely value of using AI to help management oversee and coordinate everything.

AI-aided systems can significantly improve how humans process and control many constantly changing data streams, both for adherence to plans and for detecting deviations from plans.

Deciding whether such systems are “bad” or “good” is not my focus here. I am already a firm believer in using AI tools intelligently. Instead, my focus is the potential impact of such systems on how complex processes are managed, given one simple and annoying fact: our inability to predict the future.

Unexpected things happen. Things break. People get sick and die. Accidents and natural disasters occur. The weather — or the competition, or the enemy — does something unexpected.

Sure, an AI-supported system might be able to respond to such events more quickly. But as someone with years of experience, I would question how much autonomy to give such a system in controlling how it responds to the unexpected. For example, we all know that mistakes and even disasters still occur in space exploration, no matter how many safeguards, workarounds, backup systems, and procedures we build in to manage the unexpected.

I like what Sangeet Paul Choudary says in his paywalled Harvard Business Review article, “AI’s Big Payoff Is Coordination, Not Automation”:

Over the past decade, my work has focused on designing platforms to organize and coordinate fragmented systems. But until recently, that kind of coordination had a hard limit: It depended on highly structured data and clean interfaces, which required consensus on standards and workflows. In practice, most systems varied too widely to align, so coordination remained manual and expensive. But AI’s translation capabilities have removed that limit. The new kinds of coordination that it enables are changing not just workflows but also the nature of value creation and competition in fragmented industries.

As Choudary suggests, AI-based “translation” can make obsolete the old concept of systems integration via controlled and standardized interfaces, which are difficult and expensive to maintain.

I say “can make,” not “will make.” We have to factor in the potentially significant transition costs of moving to AI-dependent system and process coordination. I’m thinking here of my own experiences with moving toward processes that require coordination and transition across traditional departmental boundaries. The politics of such changes can be fierce, especially for someone trying to sell technology-based changes into such an environment when the actual financial benefits are not immediately apparent — or cannot be accurately predicted.

That’s not to say that some organizations are not taking advantage of the transformative nature of AI-based systems, as discussed in How AI Is Changing Government Procurement. Government procurement may have a reputation for being old-fashioned and stodgy, but some government agencies are recognizing the potential for what Choudary describes.

Copyright © 2026 by Dennis D. McDonald. Image created with assistance from ChatGPT Plus Image Generator (OpenAI).

In my consulting I often use artificial intelligence tools to support the business development efforts of small and mid-size tech firms, I’m seeing three major ways AI is changing government IT procurement:

1. Changes in how IT products and services are procured.

2. Changes in the IT products and services procured by government agencies.

3. Changes in how vendors respond to #1 and #2.

1. Changes in how IT is procured

AI is changing how IT products and services are procured in several ways.

• Procurement systems themselves are being modified to incorporate AI tools, one example being the use of dedicated AI chatbots to help users interact with procurement systems. Adam Stone writes about examples in AI in Procurement Is a Game Changer for Government.

• Procurement officials are using AI tools to help research and define requirements for IT-related products and services in ways that reflect the policies of state and local government agencies. Examples of this are discussed in the Carnegie Endowment’s How Cities Use the Power of Public Procurement for Responsible AI.

One of the more significant changes I have seen while monitoring government markets for IT services is how there appears to have been a rise in the use of RFIs – requests for information -- and “sources sought” announcements compared to the total number of procurement announcements. State, local, and to some extent Federal agencies are using more  RFIs to seek advice from vendors.

Perhaps many agencies are realizing that AI-related tools are potentially more transformative than prior technology innovations. Rather than simple replacements or upgrades that process data, these potential “process disruptions” can actually reconfigure how agencies work.

The rise of AI may also be increasing use of RFIs prior to formal RFPs in this market as a way to ask vendors to speculate on how their AI-based offerings might benefit the issuing agency. In effect, agencies may be asking vendors for “free” consulting, but this also provides an opportunity for vendors to influence upcoming procurements to their advantage.

2. Changes to what is procured

Some government agencies appear seriously interested in how comprehensive and transformative AI-related services can become.

On the Federal side, for example, Department of Defense agency announcements in SAM.gov appear to be much more sophisticated analytically than those announced by civilian agencies.

At the state and local level, there appears to be an increasing number of announcements recognizing that AI support can extend beyond legacy system replacement to support for cross-system and cross-department process improvement.

Such changes call to mind the background for the traditional category of “systems integration” projects where consultants managed changes to multiple systems in ways that demanded prior understanding of existing system operations, IT functionality, business processes, management priorities, and hands-on familiarity with relevant data.

Given the transformative potential for AI tools to change and hopefully improve how people interact with complex systems, I suspect that some of these more traditional systems integration capabilities will show up as requirements in the more sophisticated RFPs now emerging.

While the era of huge “big bang” transformation projects may be over given their cost and complexity, I think it’s also possible that intelligent use of AI tools might facilitate management and control of larger system-and-process change projects. An example of how AI tools can help in managing projects with complex workflows is described in the Harvard Business Review article AI’s Big Payoff Is Coordination, Not Automation.

3. Changes to how vendors respond

How AI changes to how vendors respond to opportunities announced by government agencies requiring AI-supported tools will depend on the services to be offered. Government agencies are seeking:

1. Consulting and advice on AI transformation and IT modernization

2. Design and development of systems that incorporate AI

3. Implementation, training, and support for AI-enabled systems

4. All of the above

Given the risk-averse nature of much government IT contracting, my guess is that the market is initially likely to go to established vendors who already have market position with agency purchasing organizations as well as those who are already incorporating AI support into their offerings.

What this simplistic analysis does not take into account is the potential AI has for altering the “balance of power” in contractor supplied IT  services. Areas impacted by AI application include coding, testing, market research, project management and administration, proposal development, and ongoing support. Can smaller (and potentially more nimble) firms employ AI to at least partially overcome the market advantage of larger more established IT contractors? I think so.

Thousands of purchasing systems exist in the US at all levels of government that announce business opportunities every day. Keeping up with potential opportunities has always been challenge especially when focusing on state and local IT opportunities. Many of these local and state procurement systems are already implementing AI support for how opportunities are summarized and how they can be searched by potential suppliers.

Depending upon your market focus, a small or mid-size IT vendor might find it necessary to register a company profile with multiple systems, with each system having its own process for vendor profiles, announcements, searching, and document access. Even systems that aggregate and announce opportunities from multiple systems can still require interacting not only with the aggregator but with the original issuing agency itself. This can be a frustrating process where inevitably one finds oneself reviewing many different retrieved opportunity descriptions to find just one to forward to management for their consideration.

I have found AI support to be critical for dealing with the variability across such  systems. They all seem to have their own variations in how they interact with vendor service profiling and “standardized” commodity codes such as NAICS, NIGP, and UNSPSC. I have also found that “crosswalks” across commodity codes and different keyword combinations can be significantly helped through use of AI tools.

Conclusion

Just because AI is impacting all aspects of how government agencies procure IT services does not necessarily change the fundamentals of how individual vendors should behave. They must still discover opportunities, decide which to pursue, write proposals, and negotiate profitable contracts.

Vendors still must understand customer expectations, build relationships, and respond to vendor processes and market constraints.

At each step of the way, AI is enabling “more/better/faster.” How AI tools are employed to support business development must be carefully controlled, and this starts with an understanding of what you are selling and how flexible you are willing to be in selling it.

Copyright © 2026 by Dennis D. McDonald. Image created with assistance from ChatGPT Plus Image Generator (OpenAI).

I have always believed that technology can be used for both good and bad purposes. This includes the technologies underlying both the Internet and AI.

Separating the good from the bad when it comes to these technologies can be problematic. Reasons for this are expressed in the article by Nathalie Maréchal titled “Tech Policy Is on the Front Line of Fascism vs. Democracy. Pick a Side.”

The Maréchal article points out that the confluence of economics and politics, including the growing concentration of wealth in the hands of a few, has reached a point where decisions about how technology is employed can run roughshod over traditional so-called “liberal” concerns such as privacy, respect for intellectual property, free speech, government transparency, and civil liberties in general. Maréchal quotes TechDirt’s Mike Masnick:

“…what’s happening in the US right now is some sort of weird hybrid of the kind of power grabs we’ve seen in the tech industry, combined with a more traditional collapse of democratic institutions… The story that matters most is how the dismantling of American institutions threatens everything else we cover. When the fundamental structures that enable innovation, protect civil liberties, and foster open dialogue are under attack, every other tech policy story becomes secondary.”

There is more going on here than just “the rich getting richer at the expense of the rest of us.” The wealthy are not only getting richer, they appear to be allied with (or even controlling) government policymakers and are willing to influence communication and the flow of information in ways that actively promote the erosion of free speech, civil liberties, and social justice.

I think it is overly simplistic to blame current threats to freedom and democracy directly on technologies like the Internet and AI. It’s the old story of the fox governing the henhouse: the fox can’t be expected to implement policies that will restrict his access to the henhouse if he is also controlling how rules for accessing the henhouse are made.

The same goes for controlling misinformation, as discussed in my article “AI, Misinformation, and Trust — the Scarce Commodity.” If those controlling the Internet and AI feel disadvantaged by the truth, they will not support rules and policies designed to ensure truthfulness. This resistance to external rules and oversight also extends to resisting accountability for harms to health or safety they are accused of enabling.

Do I agree with Maréchal’s view that we have to “pick a side” in the fight between authoritarianism and democracy?

In theory, I do agree — especially here in the US where authoritarianism is clearly on the rise, as shown by the steady erosion of voting and other rights. At the same time, it’s too simplistic to attack the forces behind the Internet and AI without also acknowledging the good these tools can support. The forces aligned against democracy and the rule of law here in the US are formidable, and these forces are aggressive in how they employ the Internet and AI—much of which they control—to promote their interests.

One galling aspect of this is that we are forced to consider using methods that previously would have been almost unthinkable in the defense of democracy, such as gerrymandering election districts to counter gerrymandering in other districts.

We must also face an uncomfortable truth: there are many who appear indifferent to the erosion of democracy. One irony of this is that the same Internet and AI tools that are used by those attacking democracy can also be used to convince voters of the need to preserve democracy. This means that today’s fragmented media environment can be employed both to attack democracy and to defend it.

Copyright (c) 2026 by Dennis D. McDonald. Image created using the AI tool Coloring Book Hero.

A recent article in Science magazine, The Misinformation Accelerator, describes how generative AI is transforming the creation of misinformation.

In one case, researchers observed a disinformation site dramatically increase its output after switching to AI, building what one researcher described as a “bigger, better forest” of content to hide misleading narratives.

At the same time, controlled studies have not shown AI-generated propaganda to be more persuasive than human-written content. Researchers also continue to struggle to demonstrate that misinformation consistently changes people’s attitudes or behavior.

Perhaps focusing on the use of AI to generate propaganda and misinformation is shortsighted. I’m reminded of what happened when “Web 2.0” and social media emerged. Back then, I naively believed that lowering barriers to communication would bring people closer together. The logic seemed straightforward: if information flows more freely, understanding should improve.

That’s not necessarily what happened. I began writing about the decay of social media as advertising and political misinformation increasingly dominated these platforms. Information flows have become more fragmented. Online groups have grown more insulated. The same tools that made sharing information easier have also made it easier to reinforce existing beliefs—along with biases and outright falsehoods.

What I took from that experience is simple: making information easier to create and distribute does not necessarily improve how people decide what to believe.

Many years ago in the United States, most people received their news from a small number of television networks, along with major newspapers and magazines. Figures like Walter Cronkite occupied a central role in a more constrained information environment.

That model has steadily eroded. With so many voices competing for our attention, which ones do we choose to trust?

My personal approach is to seek out a wide variety of sources, particularly since I no longer fully trust traditional media such as The Washington Post. By scanning as many sources as possible, I hope to increase the likelihood of uncovering the truth about what is happening in the world.

One of the more important shifts described in the Science article is that researchers are beginning to focus less on the content of misinformation—which can now be created at the drop of a hat—and more on who is spreading it and how they behave.

That shift matters. The list of news sources I follow has evolved over time as I learn more about who is trustworthy, who can be relied upon, and who might be held accountable for telling—or not telling—the truth.

Technical solutions—detection tools, AI countermeasures, and platform policies—may help at the margins, but they cannot solve the trust problem. That is something each of us must work through individually.

Social media platforms such as Bluesky—and before it, Twitter—can unfortunately be significant sources of misinformation, given how easy it is to post short snippets of realistic-looking news without a source or verification.

If the past 20 years have taught me anything, it is this: making information easier to create and share does not bring us closer to the truth. It can, however, fundamentally change how we decide whom to believe.

Copyright © 2026 by Dennis D. McDonald

The Science article “AIs can ‘memorize’ data they shouldn’t. Can they be forced to forget?” describes newly available open source software designed to study how large language models can be tested for their ability to “forget” what they have learned.

Why should an LLM be able to “forget” what it has learned? Several reasons:

• training input later found to be inaccurate

• potential for release of private or personal information

• data output too closely copying or mimicking protected intellectual property

An important question is whether it is even possible for LLMs to ever forget everything they are trained on. While tools might be able to test whether targeted data have actually been removed, I can imagine that the resource cost of such removal/forgetting would be substantial, especially if it has to be done on a regular basis and also addresses any downstream copying or ripple effects occurring since introduction of the targeted data.

Predicting impacts of such “forgetting”—and the potential costs of impacted decisions made or actions taken before the forgetting took place—could be substantial and perhaps folded into the ongoing maintenance costs of the LLM.

This raises some legal issues of liability that might keep the lawyers busy and well fed, but they could be real nuisances for users who base decisions on outdated, illegal, or inaccurate output.

One might logically ask if this question is any different from what any question-answering system—or any medium for that matter—can be held “liable” for actions taken or decisions made based on use of that medium, regardless of the medium in question.

That question is beyond the scope of this piece. We already know that there are businesses devoted to expunging inaccurate or embarrassing information from the web. I would not be surprised if similar entities emerge to address the training and output of LLMs; again, this is an issue that will most likely keep the lawyers busy for years to come.

Copyright (c) 2026 by Dennis D. McDonald

More on “Accountability”

On the difference between illustrating reality and evoking ideas

My initial reaction upon reading Why Science Communication Needs Visuals was straightforward: “Yes—but what if the scientist is not an artist?”

The reasons given by the Association of Science Communicators article for incorporating art are clear enough:

1. Capture attention in an information-saturated world

2. Explain complex phenomena

3. Take advantage of the transformative power of conceptual metaphors

The first two are familiar. “Capture attention” is a practical necessity when addressing a general audience, and “a picture is worth a thousand words” has long served as shorthand for the explanatory power of images.

The third point—using conceptual metaphors—is where things become more complicated.

To begin, an author must decide which idea is actually worth communicating. In any piece containing multiple ideas, that can be difficult.

Once that choice is made, the next challenge is finding a visual metaphor that helps convey the idea clearly. This is possible, but not always easy—especially when the metaphors that first come to mind are predictable or overused. (Robots representing AI is a device I have used myself. I apologize!)

At this point, it is worth making a distinction that the article tends to blur. An artist can certainly help make a complex event, action, or idea understandable. That is one kind of contribution—grounded in clarifying or representing physical reality, even when the underlying phenomena are not directly observable. But that is quite different from representing a concept metaphorically or symbolically. These are two distinct kinds of artistic applications, and they serve different purposes.

A scientist reporting a research finding would, I suspect, generally prefer the former: a visual representation that reflects physical reality as faithfully as possible. That kind of illustration supports understanding of what actually occurs. A metaphorical or symbolic image, by contrast, does something else—it frames how we think about the idea rather than showing what it is.

The authors of Why Science Communication Needs Visuals acknowledge the difficulty faced by non-artists and caution against overreliance on AI-generated imagery:

“AI tools can be incredibly useful to artists (I admit to using them myself), but the tool cannot replace the artisan. Because science communication needs more than just visuals—it needs the human creativity, understanding, and skills that only trained illustrators can provide.”

I’m sympathetic to that view. I’m not an artist, but I regularly need images to accompany the articles I publish in the Managing Technology section of my website. When I cannot use my own photographs or readily available sources, I turn to tools like ChatGPT to generate images.

What I am usually looking for is not an illustration of a specific event—say, what happens when two molecules interact—but rather a representation of a key idea. The images I use are often metaphorical rather than descriptive.

Producing those images is rarely straightforward. It typically involves a back-and-forth process in which successive versions are refined through prompt adjustments. Over time, I have learned how to guide that process more effectively, though it remains something of an iterative negotiation between intention and outcome.

Does this place me in the camp of those who see AI tools as replacements for human artists?

Not really. I have hired artists in the past for professional or commercial work, and I would do so again where the stakes justify it. But for the short, exploratory pieces I publish on my personal website, that has never been a practical option. In that context, AI-generated images are sufficient.

But I would still prefer to work with a professional artist whenever fidelity to physical reality matters. I am not trying to document phenomena so much as to suggest ways of thinking about them.

That distinction matters. There is a meaningful difference between illustrating a phenomenon and evoking a concept. The former aims at accuracy; the latter at interpretation. Both have value.

For now, I continue to experiment with tools that allow me to represent ideas more economically, if not more precisely. What I am learning is that conceptual images—while imperfect—can still be useful.

Copyright © 2026 by Dennis D. McDonald

Addendum 1: Examples illustrating the distinction discussed above:

• Conceptual illustration that works: https://www.ddmcd.com/managing-technology/impacts

• Conceptual illustration that falls short: https://www.ddmcd.com/managing-technology/mynews4

Addendum 2: ChatGPT’s own reverse engineered prompt describing generation of the artist-at-his-easel image at the head of this article:

A thoughtful, realistic scene of an artist standing at an easel inside a large, complex scientific environment, such as a particle accelerator or advanced laboratory. The surrounding machinery—pipes, cables, control panels, and structural elements—is rendered in monochrome (black, white, and shades of gray), highly detailed and precise, emphasizing physical reality and technical complexity.

The artist is actively painting on a canvas. In contrast to the monochrome environment, the painting is vivid, colorful, and expressive—an abstract, symbolic interpretation of what the artist sees. The canvas contains flowing shapes, light, and conceptual imagery (such as energy patterns, connections, or abstract forms), not a literal depiction of the machinery.

The composition should emphasize the contrast between the grayscale “real world” and the colorful “interpreted world” on the canvas. Lighting and composition should draw the viewer’s eye to the painting while still showing the scale and complexity of the machinery behind the artist. The overall tone should be thoughtful, reflective, and slightly cinematic, suitable for an essay about science, art, and interpretation.

What recent experiments with large language models reveal—and what they may never fully replace in the practice of science.

Testing Whether AI Can Do Science

Celina Zhao’s report in the Feb. 27, 2026 issue of Science, How will we know if AI is smart enough to do science, describes tests to determine how large language models can help scientific research. The tests described range from very detailed fact-checking to simulations that model how scientists formulate and test hypotheses.

It’s likely that such tests—initially encouraging—will become more robust, detailed, and realistic over time. Whether they can actually replace human scientists is another matter and raises questions about creativity and human thinking processes.

There is little doubt that much of the grunt work associated with data collection and analysis can be enhanced and streamlined by AI, as can the retrieval and interpretation of details from the published literature. In AI, journals, and the evolving knowledge divide I discussed how AI-based tools are being introduced that can interact intelligently with scientific literature. It’s likely that such tools will evolve rapidly. Still, several questions deserve careful consideration when discussing the role of AI in scientific research.

Question 1: Access to Data

The first concerns data access. If an AI tool is used by a researcher at any stage of a research project, will access to the published literature be sufficient, or will access to the underlying data that supported that research also be required?

There is no simple answer to this question, given the complexities associated with accessing and interpreting original research data. One issue involves how easily an AI agent might obtain and analyze such underlying research data, as well as the cost and resources required to do so. In theory, an intelligent AI system might seek access to original data if, for example, it detects inconsistencies or possible errors in what it is reviewing. Yet the logistics, cost, and resources involved—especially if the data reside in proprietary systems or involve human subjects—could prove problematic and require human intervention in the process.

Question 2: Creativity and the “A-ha Moment”

The second issue associated with AI involvement in research concerns creativity—how we define it and what level of creativity we expect an AI system to exhibit. Identifying hidden similarities or relationships is not the same thing as creating something genuinely new. Can AI assist in the creative processes involved in scientific research? The answer is almost certainly yes. But can AI generate something truly novel—something we might describe as an “a-ha moment,” a sudden insight or realization? I’m less certain.

Question 3: Serendipity in Scientific Discovery

The third concern relates to serendipity and the role it plays in scientific discovery. I first encountered this when I was a graduate student interviewing scientists about how they located and used articles in scientific journals. Some were adamant they did not want to rely solely on stored computerized searches or narrowly defined logical queries. They wanted exposure to research outside their immediate focus areas, precisely because they might encounter something from an “unrelated” field that could ultimately prove relevant to their own work.

Could an AI agent be programmed to scan unrelated areas of research in the hope of finding something useful for a particular researcher’s interest? It’s plausible that an AI system could be “fine-tuned” to look beyond a researcher’s primary focus area in search of potentially relevant insights and then learn from human feedback which discoveries were actually useful. Perhaps such an approach, which isn’t that different from scanning related areas out of curiosity, should then be called “planned serendipity.”

Who Benefits First?

When all is said and done, there is little doubt that AI can help—and already is helping—researchers around the world. One group that may benefit especially in the near term is graduate students, who have long played a central role in research projects by carrying out much of the behind-the-scenes work that scientific progress requires. When this happens, the educational process will also need to include how best to manage AI in research work.

Copyright © 2026 by Dennis D. McDonald

Many years ago I worked on a US National Science Foundation project to enhance researcher access in a developing country to scientific literature. I focused on academic libraries. Other team members focused on accessing online bibliographic databases via leased satellite channels.

For a young US researcher that project was a real eye opener given the wide information access we had long taken for granted in the West, even in those pre-internet times.

What made a lasting impression on me was coming face to face with the lengths some students and researchers had to go to get the information they needed for their research and study.

In one case I interviewed a local college professor who had a side business tutoring his students. He literally rented out photocopies of articles he had brought back from Western libraries to students who attended his evening seminars. Even there a distinction had developed between the haves and have nots: if you couldn’t afford the evening seminars and the rental price for reading a journal article photocopy, you were out of luck.

Reading Jeffrey Brainard’s article “Journal giant Elsevier unveiled an AI tool that scans millions of paywalled papers. Is it worth it?” in Feb. 20, 2026 Science, I was reminded of that long-ago project and how wide the gulf is for some in their quest to access scientific literature.

Brainard describes publishing giant Elsevier’s new tool LeapSpace, which uses a large language model (LLM) to scan a huge collection of paywalled papers to answer researcher questions. LeapSpace covers journals from Elsevier as well as those from Emerald, Institute of Physics, New England Journal of Medicine Group, and Sage Publications. The system doesn’t just provide answers; it also provides the sources for its answers so the researcher can evaluate sources and reliability.

Such AI based systems raise questions like the following:

1. Will this system replace or increase demand for original sources?

2. How will this impact what institutions and libraries already pay for journal subscriptions?

3. How will government regulations treat such access to articles based on publicly funded research?

4. What happens to knowledge published in the many open access journals not covered by the Elsevier product?

5. Will this service disadvantage researchers who are not associated with a journal-subscribing institution?

Such questions got me thinking again of that professor long ago who rented access to his photocopy collection. Based on my own experience using basic AI tools like ChatGPT for both research and analysis tasks, there is no question in my mind that, when properly managed, tools like LeapSpace can be incredibly powerful productivity boosters—for those who can effectively use them.

Addressing the above questions requires consideration of economics, ownership, intellectual property management, and politics. As suggested by Korov and Oreskes in “Lineage of Science in a Warming World: Who Owns Climate Knowledge,” such issues eventually raise the prospect of government regulation::

“In the United States, any regulation of data and models will require us to move politically uphill. But that makes it more important to be precise about what is happening. This is not just ‘data privatization’ in some abstract sense. It is the privatization of knowledge: who gets to know in detail how risks are distributed; who gets to contest those assessments; who gets to change them.”

Copyright (c) 2026 by Dennis D. McDonald

Thoughts about COPYRIGHT

I make no bones about my personal and professional use of AI tools, about which I have written frequently here. I manage the associated risks through careful and disciplined use, and I advise others to do the same.

One area that directly affects both me and my clients is the way government agencies are increasingly incorporating AI-related guidelines into IT procurements. Some agencies specifically request systems that integrate AI-based processes to streamline or accelerate workflows. Others embed definitions of AI into state laws and local policies that govern how AI may be used in public services.

None of this should come as a surprise. Government agencies have always been vigilant about controlling how personal and private data are managed. A growing ecosystem of public bodies and advisory organizations now exists to help agencies — and state legislatures — understand how emerging technologies affect data governance and public accountability.

Some may view these new rules and regulations as a hindrance to innovation; others will recognize them as an opportunity.

I fall into the latter camp.

At a minimum, any government IT contractor worth its salt understands the importance of staying current with business-relevant laws and regulations as they evolve. The real challenge is figuring out how to get ahead of these developments when it comes to AI.

If you have a solid understanding of your target market, you are likely already monitoring how procurements are being shaped not only by AI but also by cybersecurity, privacy, and data access requirements — along with the laws, regulations, and standards bodies influencing them.

I wrote about this broader governance challenge in 2024 in “Towards Unified Governance of AI, Data, and Cybersecurity Initiatives,” where I argued that AI governance cannot be separated from data stewardship and cybersecurity controls. The issues now surfacing in procurement language are, in many ways, an extension of that same conversation.

In this environment, it is no longer enough simply to have internal AI practices. You must be prepared to articulate your AI policies clearly — on your website, in promotional materials, and across public channels — and, just as importantly, in a format that can be rapidly adapted for proposals and quotes.

What we’re seeing is the normal process of institutions figuring out how to live with a new technology. Procurement language shifts. Definitions evolve. Policies adjust — sometimes awkwardly. That’s simply how these cycles unfold.

For those of us who work in and around public-sector technology, the question isn’t whether these rules will appear, but how thoughtfully we respond to them. Paying attention early, and preparing accordingly, is far more productive than reacting later.

Copyright © 2026 by Dennis D. McDonald

It’s getting harder to keep up with the news these days. I’ve been tracking the sources I’ve used over the years (e.g., see Where I Get My News These Days). My sources, paid as well as free, have expanded even as some sources have fallen by the wayside.

I currently see two major forces at work that threaten the public’s ability to keep up with current affairs.

The first is the censoring of legacy as well as modern media by corporate owners, CBS News and The Washington Post being two premier examples. The deterioration of the latter has been especially painful given my having subscribed for over 30 years.

The second is the increasing difficulty of tracing sources for news that spreads through social media. An example is the social media platform Bluesky, where it is common to see news snippets promoted without any attempt to cite a source. Asking for a source—even for items that I would be inclined to agree with—still often generates a response of “just google it,” which I’ve criticized before (e.g., see Don't Just Say "Google It!" Part 2).

This list changes frequently. I certainly don’t make any claim that it’s perfect. I’ve simply found that, when it comes to keeping up with what’s happening, seeking out multiple sources is the best way to go, even if some sources are personally distasteful.

Media

• All About Jazz Music, Musicians, Bands & Albums

• Asian Movie Pulse Reviews News Features Lists for Asian Cinema

• critical popcorn

• Far Out Magazine | Music, Film, TV, Art & Pop Culture News

• F*XOXO – FXOXO.COM

• GoSeeTalk

• InBetweenDrafts

• Salon

• SmartBrief on Media & Entertainment

• UPROXX – The Culture Of Now

  Science & Tech

• AAAS Science News

• Electric Vehicle News

• Emerging Tech – MeriTalk

• Gizmodo

• MedPage Today

• NASA

• Physics World

• The Few Shot Prompting Guide

• Scientific American

• Smithsonian

• Space.com

• SpaceNews

• TechCrunch | Startup and Technology News

• Technology - Enterprise Viewpoint

• Tech Policy Press - Technology and Democracy

• Science History Institute

Politics & Government

• ARPA-H News

• The Contrarian

• Defense One

• Democracy Docket Premium

• Lawfare | Lawfare

• The New Republic

• NIST AI Engagement | NIST

• Popular Information

• Radio Free America – Heartland Signal | Radio Free America

• Senator Warner

• Senator Kaine

• Congressman Beyer

News News

• AP News

• Axios

• BBC News

• The Economist

• Guardian

• New York Times

• NPR

• News From The States.

• The Onion

• ProPublica

• Reuters

• Reuters ROI

• Semafor

• States Newsroom

World

• Al Jazeera English

• IndiaToday

• The Kyiv Independent  

• Italia Mia

• Mainichi (Japan)

• South China Morning Post

Local

• ALXnow - Alexandria Now

• Connection Newspapers

• insidenova.com  

• Patch Alexandria

Virginia

• bluevirginia.us

• (1) Blue Virginia Breaking News | Substack

• Virginia Mercury

• VaNews

• Cardinal News

Social

• Bluesky

• Reddit

• LinkedIn

• Substack

• Truth Social

Copyright (c) 2026 by Dennis D. McDonald

More about THE MEDIA

One of the first things you learn when you become an independent consultant is that you are largely on your own when computer problems arise. Over time, I’ve watched that reality shift—from phone based call centers or or inhouse staff to ChatGPT.

My current use of ChatGPT (I pay for a subscription, which remembers my conversations) has helped resolve problems related to printers, software installation, cameras, microphone controls, non-responsive online help systems, and phone–computer integration. I’ve even used ChatGPT to categorize text data and clean up lists that later served as mail-merge inputs for promoting my consulting services.

ChatGPT (and, I assume, other AI tools) combine several powerful features:

1. A generalized diagnostic approach that adapts to different situations (for example, hardware versus software, installation versus day-to-day use, etc.)

2. Conversational information exchange, rather than forcing users through rigid, pre-structured forms.

3. An evolving repair process, where the solution and discussion evolve as more detail is added.

4. Persistent conversations that can be paused and resumed later with no loss of context—a real boon when multitasking.

5. Multiple input modes, allowing information to be shared via text, voice, or images.

Sure, it would be nice to have a personal tech-support specialist who knows my name. But I can contact ChatGPT at any time of day, and we simply pick up where we left off.

Let’s be honest, though: dealing with traditional tech support or with ChatGPT—whether through a live human, a phone call, or a keyboard—requires time that can be more productively spent on income-generating work. For now, though, I’m genuinely glad I have ChatGPT and don’t have to rely solely on my own ignorance!

Copyright © 2026 by Dennis D. McDonald.

Why Experience, Judgment, and User Roles Matter in Intelligent Systems

It started with a letter

This article began as a handwritten letter to an old friend—a former client now retired—with whom I share long-standing personal and professional interests. I wanted to give him a candid perspective on my use of AI tools such as ChatGPT in my consulting including work for an IT contractor whose business development efforts I support.

What I have been encountering in my day-to-day use of AI tools for both personla and professional use raises questions about system design and, ultimately, AI governance. For example, I believe an understanding of an AI tool user’s role—including the user’s nresponsibilities, scope of authority, and decision-making context—can and should influence how that tool is designed, configured, and managed.

Before sending the letter, I scanned it and uploaded it to my ChatGPT Plus account. After confirming that the system could accurately convert cursive handwriting into editable text, I asked it to summarizee the letter’s main ideas as the basis for a website post while deliberately removing references to specific products, services, or companies.

Since then the following text has gone through several iterations as my letter’s original ideas were revised, expanded, and interpreted with the help of ChatGPT.

Experience, Judgment, and the User’s Role

Some contemporary software systems—including those incorporating AI—are designed to manage complexity by modeling and applying entire workflows or lifecycles. These systems are often comprehensive by design. They assume a sequence of activities that begins with discovery, moves through evaluation and decision-making, and continues into execution, monitoring, and documentation.

Logic behind the tool

From a design perspective, this approach is understandable. Systems built around end-to-end processes provide structure, continuity, and consistency, especially for organizations seeking repeatability and oversight. Knowing that there is a well structured logic behind what an AI tool does or suggests helps to build confidence in its interpretations and recommendations, akin to the confidence engendered by a professional that broadcasts his or her training, qualifications, or certifications.

When combined with AI-based capabilities, such systems can also operate at a scale and speed that would be impractical using manual methods alone. However, this same strength can create friction when users engage with the system from different roles and with different objectives.

One user’s focus

In my work with one consulting client, for example, I operate primarily at the front end of a complex process. I help identify business opportunities. My responsibility is not to execute an entire business development lifecycle but to help determine which opportunities warrant management’s attention. This requires broad scanning, comparison, and prioritization of potential opportunities rather than sustained involvement in downstream activities.

When a system is designed around full lifecycle participation, it may implicitly assume that all users intend to progress through every stage. Recommendations, prompts, and suggested “next steps” can be generated on that basis. For users whose role is intentionally limited, this can result in guidance that is technically sound but operationally misaligned and time-consuming to address.

Adapting to individual users

This highlights a key design consideration: complex AI-supported systems may need to adapt not only to different organizations, but to different users within those organizations. A junior analyst, a subject matter expert, and a senior advisor may all rely on the same data and tools, yet their goals, constraints, and measures of effectiveness differ. Systems that recognize and accommodate these differences are more likely to support effective decision-making. They also support more efficient resource allocation, since time is money for any organization concerned with controlling costs.

One area where AI-enabled systems can excel is their ability to learn. Tools that remember prior searches, decisions, stated preferences, and patterns of interaction can significantly reduce repetitive work and improve relevance. Such embedded learning and memory features allow the system to adapt to how a user actually works and how that user’s responsibilities are defined.

It is also important to distinguish between tasks that benefit most from automation and those that do not. AI is exceptionally effective at large-scale research, information retrieval, and pattern detection. When guided effectively, these capabilities can significantly reduce the time and effort required to assemble context and options. Given my own experience managing market research for technology-based systems, I continue to be amazed at how AI-based search tools accelerate market and competitor analysis.

Informing judgement

Mature judgment, however, operates at a different level. Decisions about priority, risk, and strategic fit are shaped not only by information, but by experience, organizational knowledge, and an understanding of strategy and consequences. In my view, the most productive use of intelligent systems is not to replace judgment, but to support it—by handling repetitive and computationally intensive tasks while leaving role-specific decision-making in human hands.

As systems continue to evolve, their effectiveness will depend less on how much process they incorporate and more on how well they adapt to varied modes of use. The ability to adapt may ultimately matter more than attempting complete automation of a process, particularly when that adaptation takes into account awareness of the user’s role and responsibilities.

From Role Awareness to AI Governance

The observations above naturally extend beyond system usability into questions of governance. If AI-enabled tools adapt to users and their roles over time—learning preferences, remembering decisions, and shaping what is presented next—then how those adaptations are guided, constrained, and reviewed becomes an organizational concern.

Multiple pathways

Software developers have long recognized that “one size fits all” is rarely appropriate for complex systems. Today’s tools can learn about user preferences, experience levels, and past behavior, enabling them to better anticipate expectations and support decision-making when users return.

What is new is the scale and influence of modern AI systems. These systems often exceed the experience of individual users not only in the volume of information they process, but also in their ability to analyze and determine what information is surfaced and how it is packaged. Those that incorporate embedded process or lifecycle models can shape both what the user sees and what the system encourages the user to do next.

Experience and context

This is where user experience and organizational context become critical. As noted earlier, one of my own roles involves identifying potential business opportunities for one client across multiple technology domains. How I interact with AI tools has been shaped shaped by decades of consulting, research, project management, and corporate management experience. This experience informs how I assess relevance, risk, and priority when reviewing a list of search-retrieved opportunities and deciding which to forward to my client for consideration.

At the same time, the analytical capacity of AI systems far exceeds what any individual can match. Research tasks that might take me a full day can often be completed in seconds. This significantly accelerates parts of the research process, but it does not eliminate the need for judgement and collaborative decision-making. Decisions about whether to pursue an opportunity still require management judgment—particularly when considering proposal effort, pricing strategy, staffing, and partner engagement.

This brings us back to role awareness. For AI tools to be effective and responsibly deployed, they must account for the user’s responsibilities, experience level, and scope of authority, as well as how that user’s work is reviewed and acted upon within the organization. Ideally, the system learns about both the user and the organizational context in which the user operates and adapts its behavior accordingly.

This is where AI governance enters the picture. Managing how AI tools are used begins with understanding what they are intended to support and how they align with organizational goals. Traditional consulting engagements typically start by developing an understanding of a client’s business strategy and operating model. The same principle applies to AI systems.

One possible approach to facilitating AI governance in such situations might be creation of an evolving, two-part profile: one describing the business or organizational context in which the tool is used, and another describing the roles, responsibilities, and experience levels of its users. These profiles should be created, reviewed, and updated over time and made available to both users and management.

Alignment with organizational priorities

Doing so increases the likelihood that adaptive AI systems remain aligned with organizational priorities rather than simply optimizing for an externally designed process’s completeness or technical capability.

In that sense, this article’s origins—as a personal letter reflecting on everyday work—are not incidental. Questions of AI governance often surface first at the point where real people, real roles, and real responsibilities intersect with increasingly capable systems.

Governance must adapt

How governance operates in this context will necessarily differ from organization to organization. Rigid, hierarchically structured organizations will govern AI use differently from more collaborative organizations where departmental boundaries are less distinct.

Another difference lies in whether AI tools are viewed as traditional, purpose-designed applications that support consistent execution of repetitive processes, or as embedded “advisors” tuned to individual preferences and informed decision-making. Either way, failing to consider how AI tools should be governed is a major mistake.

Copyright 2025 by Dennis D. McDonald, Ph.D. This article was written with the aid and support of ChatGPT, which acted as editor, advisor, and—in some cases—collaborator. While the core ideas are my own and originate in a personal letter, ChatGPT helped frame and refine the discussion through interaction over several days in December 2025.

My Use of AI Tools

I’ve made no secret of my use of AI tools—here on my website, in personal matters (for example, repairing home appliances), and in my consulting work. I pay for a ChatGPT Plus subscription, and I frequently use AI-based tools such as GovTribe to support client business development.

Writing and Editing Support

The closest I’ve come to using AI’s “creative” abilities is in the development of some of the graphics that accompany the Managing Technology section of this website, such as the graphic displayed at the top of this article. I do often use ChatGPT Plus to edit the text of the articles here—for clarity, grammar, and spelling—after I’ve written out the text of an article in longhand and used speech-to-text software on my iPhone to create an editable file. I did that with this article and it did make some of my awkward sentence structures more straightforward.

I have “trained” ChatGPT not to be too creative in its edits. In the early days, the edited text sometimes didn’t sound like something I would have written. That has changed. I do have a habit of writing overly long sentences and of writing in cursive with a fountain pen, and ChatGPT’s editing has helped me rein that in.

Disclosure?

But the question remains: even if I mention when I use ChatGPT to create a graphic header image, should I also mention when I use ChatGPT—or any other AI tool—to edit text for clarity, grammar, and spelling? Should that usage be disclosed in the accompanying text?

My View

I don’t think so even though I’ve done that here. I’m not copying the likeness of a dead celebrity, nor am I trying to make my writing sound like Hemingway, Steinbeck, or Arendt. Still, efforts are underway around the world to require content providers to disclose their use of AI and to manage intellectual-property controls. This probably won’t apply to a self-published website like www.ddmcd.com, but who knows what the future holds?

Practical Use

I don’t have the same concerns when using a tool like ChatGPT to repair home appliances or domestic electronics. There I can get detailed instructions on problem diagnosis and remediation, often adapted to my specific situation. I do wonder, though, where such information comes from—especially step-by-step instructions that address the quirks of obsolete models and makes.

In those cases, the proof is in the pudding: did the step-by-step instructions for installing Ubuntu Linux on a vintage 2009 iMac work or not? If there were “hallucinations” along the way, the end result was still successful and I’m back to having a fully functioning PC in the basement.

Low Stakes and the Limits of Law

Presumably, most of these instructions are gleaned from publicly available information. I do understand how using proprietary information without attribution can be problematic, and I’ve long considered the copyright notice at the bottom of my articles to be mostly pro forma—insufficient to ward off intellectual-property predators.

I learned long ago that the fuzzy nature of “fair use” would never be sufficient in the face of today’s broad and fluid access to information, but I figure that with a website like this, which I’ve been publishing on since 2002, usage is impossible to control once I hit the “publish” button.

Perhaps I would feel differently were I treating this website as something other than a hobby and self-promotion effort based on my enjoyment of writing. For all I know, somewhere an AI program is skimming my book reviews for information about historical novels set in ancient Rome. That’s fine with me—as long as it mentions my name!

But I doubt that will happen, even in countries with legislation requiring AI programs to acknowledge individual sources, regardless of the source;s published copyright or reuse notices.

Public Data

The issue of attribution becomes different when AI is used to interact with an explicitly public body of knowledge. In my business-development consulting, for example, I use tools like GovTribe and ChatGPT to research IT-related opportunities across federal, state, county, and local government. GovTribe, for example, provides a sophisticated chat-based approach to researching and managing government business opportunities—most of which are, by definition, in the public domain. I’ve also found that ChatGPT’s search and reasoning capabilities significantly enhance the market research one person can do.

Transparency and Responsibility

Both services are explicit about the steps they follow and the sources they use. As someone who spent many years in contract research, both services go a long way toward easing my concerns about what sources are being used to answer specific questions. I still have to evaluate the value of the information they provide, of course, and that’s where my own project management and consulting expereince comes in handy.

Human Analogy

Which brings us back to acknowledging the role of AI in one’s professional or creative endeavors. How different is it, for example, to use an AI search tool to solve a critical technical problem compared with walking down the hall to ask an esteemed colleague for advice? (Since I usually work by myself from my home office there’s no colleague “down the hall” but hopefully you get the point!)

If that colleague’s advice is critical to your research, professional courtesy suggests you acknowledge their contribution when reporting the results. Shouldn’t you do the same if the critical advice is provided by ChatGPT, Watson, or an AI tool focused on data modeling or intelligently processing large volumes of unstructured text?

Academic Debate and Everyday AI

We know there is ongoing controversy about the role of AI in all phases of academic research and publishing—from hypothesis formulation to data analysis to writing, editing, and peer review. Should such tools always be reported?

As AI becomes embedded in everyday domestic, professional, and recreational pursuits, I doubt that public acknowledgment of its contributions can be effectively legislated—except perhaps in high-visibility, high-value domains such as media and entertainment.

One thing we’ve learned in the past is that attempts to legislate how technology can or cannot be used inevitably become outdated or are circumvented, often controlled by those with the deepest pockets.

Tech Regulation and AI Literacy

Personally, I’d rather devote my time and energy to intelligently managing the use of these tools—and that includes teaching children how to use AI responsibly. Doing so starts with skepticism and curiosity. And before I trust a recommendation or finding from an AI tool, I want to understand where it came from!

Text copyright (c) 2025 by Dennis D. McDonald, who freely admits ChatGPT helped him edit and illustrate this article!

Back in 2007, I published a blog post titled Using New Media To Sell Old Media. It described the chain of events that led from hearing about the book American Born Chinese to buying a physical copy of that book.

My wife and I recently had a another book-buying experience that led to our buying a physical book, and I thought I would describe it here for comparison.

My granddaughter is in second grade and has started reading. We wanted to get her started on a mystery series for kids. Here is the sequence of events:

1. Ask ChatGPT for a curated list of recommended mysteries for second graders.

2. Look at the Barnes & Noble website for the first in a series for books currently available at the Barnes & Noble store near us.

3. Go to that store, examine the books, and purchase one. We purchased “A-to-Z Mysteries Book One, The Absent Author,” by Ron Roy.

4. We gave the book to our granddaughter.

5. Later that night we received a video on our iPhones from her parents of her reading the book out loud—and smiling.

Visiting the Barnes & Noble physical store was quite an experience. We are heavy readers and frequent library users, and we regularly buy physical, audio, and e-books. Being in a large and busy bookstore once again made me wonder how many others were there based on online research, and how many were there just browsing? 

But it was just a pleasure to see all the displays of books laid out for perusal. My “to be read” stack at home is already high so I had to resist buying more for myself!

The biggest difference between my 2007 expereicne and this one, of course, was the initial use of ChatGPT to locate a curated list of books to choose from, based on this prompt:

I am looking for popular mystery books that would be appropriate for 5 to 7-year-olds. It can be collections of stories or it could be series. When I was a child, I loved the Rick Brant electronic adventure series; I’m curious about whether books like that still exist. Perhaps there are more modern series like Hardy Boys that we could look at for our grandchildren?

ChatGPT came back with a long list of possible series and individual titles, and based on that recommendation we decided to go to our local Barnes & Noble to examine them physically, since the B&N website also told us which were in stock there.

I know there was once a fear that modern media would be the death of traditional publishing, and I heard that fear firsthand during my years of involvement with electronic publishing.

I don’t know how serious that fear is now. My own experience as described here shows multiple modern media supporting the purchase of a physical book. That seems pretty normal to me.

Copyright (c) 2025 by Dennis D. McDonald

Introduction

If you want to control AI, you first need to understand what you mean by control. That also means understanding what influence you have over your relationship with it. There are several ways to look at this;these are discussed below.

Discussion

1. Control Depends on Relationship Type

To talk meaningfully about “controlling” AI, we must first understand our relationship with it. Control operates differently when AI functions as a subordinate tool versus when it acts as a collaborator or advisor. Assumptions about authority, trust, and autonomy shift depending on that role.

What came to mind when considering these relationship types was how novelist Robert Harris described the evolving role of Tiro across his three novels about Cicero—Imperium, Conspirata, and Dictator. At first, Tiro was an educated slave who could read, write, and take notes. Over time, his relationship with Cicero evolved into that of a trusted advisor and eventually a free man.

In just the short time I have been using various AI tools, I’ve seen my relationship with them evolve—partly in response to what the AI learns about my interests, roles, and responsibilities. I’ve learned, for example, to be very explicit about what I want the AI tool to do, even for seemingly basic tasks such as parsing unstructured text into structured spreadsheets and databases. The more the AI understands about your background and intent, the more helpful it can be—though it can also be annoyingly overhelpful with its suggestions.

2. Roles and Relationships Evolve

AI systems already occupy a range of roles—secretary, trusted advisor, coworker, teacher, artist, or analyst. Each role carries its own expectations for reliability, creativity, and accountability. Understanding which role we’re engaging shapes how we use AI and how we evaluate its behavior.

One complicating factor is that these roles and relationships can shift instantly, which can cause confusion. Again, being explicit about one’s expectations when crafting a detailed initial prompt—including specifying what prior stored information should be consulted—is important.

3. Feedback Personalizes the Interaction

AI’s adaptive responses to user feedback can create a sense of personalization that mimics not only social interaction but also learning that influences subsequent decisions. This can make interactions feel cooperative or even intuitive. It can also mask the complexity of the system beneath.

I experienced this recently while using ChatGPT to diagnose and resolve a series of hardware and software problems—including figuring out why the inside light of my dryer worked but the control panel did not.

Being able to provide and receive feedback over the course of an interaction is essential, especially when dealing with systems or issues outside your own expertise. Understanding the steps the AI is following to solve your problem is critical. That’s why I find it helpful for the AI to provide ongoing descriptions of how it is addressing a problem. I may not understand a scrolling display of Python code, but I will understand periodic text explanations of what the AI is “thinking” as it works through an issue.

4. Collaboration Can Blur Human–Machine Boundaries

When AI assists in problem-solving, it can be difficult to separate human input from algorithmic contribution. You may find yourself wondering whether the AI’s response is (a) copied from something it read that resembles your query, (b) derived from well-understood rules or conventions, or (c) simply fabricated.

I encountered this recently when I asked an AI tool how House Atreides made its money before being sent to Arrakis. (Those familiar with the Dune universe will understand the question.) I haven’t read all the Dune books, but I assumed that somewhere there must be a description of the Atreides’ revenue sources. Still, I was surprised when one supposed source turned out to be training the Emperor’s Sardaukar terror troops. Who knew?

5. AI as Creator and Critic

Recent experiments where AI systems both author and review scientific papers show how far this partnership can go. A recent Science magazine article, At futuristic meeting, AIs took the lead in producing and reviewing all the studies, described a conference called Agents4Science that intentionally flipped conventional academic norms by making AI systems the lead authors—and the reviewers—of all presented research papers.

Of 315 papers submitted, 48 were accepted, with AI handling everything from hypothesis generation (57% of submissions) to substantial writing (90% of all papers).

Reading that article inspired me to write this one. I had already been thinking about how AI can support many stages of the research communication process, including:

• Designing and conducting research

• Drafting manuscripts

• Preparing figures, tables, and supplementary materials

• Ensuring compliance (ethics, conflicts, data sharing)

• Selecting target journals

• Formatting manuscripts

So, how “good” were the AI-generated and -reviewed papers? According to the Science article, results were mixed. But the real issue—when considering what kind of relationship we want with AI—comes down to judgment. There may be benefits such as speed, cost-effectiveness, and scalability, but judgment, expertise, transparency, and—see next section—accountability remain essential.

6. Acknowledgment and Attribution

If an AI contributes meaningfully to a task, how should that contribution be recognized? We credit human colleagues for intellectual input; should AI systems be treated merely as instruments, or as collaborators deserving acknowledgment?

This article, for example, has been developing in my mind for several months and reflects my long-standing interest in scholarly publishing and creativity. (My Ph.D. dissertation, many years ago, involved developing a mathematical model of how astrophysicists and cancer researchers select the journals in which they publish.)

The actual writing began with ten handwritten pages of notes about control, AI, and relationships. I converted those notes to text and fed them into ChatGPT with the prompt:

The following is a stream-of-consciousness set of notes about my view that controlling AI depends at least partly on the type of relationship we want to have with it, and this will be impacted by the complicating factor that AI can play so many different roles. Please summarize this into 6–10 key points for further writing.

This initiated a back-and-forth where ChatGPT not only listed key points but also—without my asking—drafted ideas for discussing them. While I often use ChatGPT to edit this website’s text (“please edit the following text for grammar, spelling, and clarity”), I rarely use it to generate text. I sensed that, if allowed, ChatGPT would have written this entire article.

Which brings us to the question of acknowledgment and attribution. Should the fact that a researcher uses AI tools at various stages of the process always be made public—just as a colleague’s contribution might be acknowledged in a published paper?

For some tasks, disclosure would be like reporting the statistical package used to calculate basic descriptive statistics. But for others—such as developing an original hypothesis—shouldn’t that involvement be disclosed?

Conclusions

The role of AI in research and communication is still evolving. Until we have a better sense of how to control and relate to its many different uses, transparency about how AI is used will be essential. This will be just as important for research published in academic journals as for the images and news carried by legacy and social media.

Copyright © 2025 by Dennis D. McDonald. Concept and final composition of the accompanying image was created through an interactive exchange with ChatGPT (GPT-5), exploring variations on light, form, and symbolism to illustrate the balance between control and automation.

As AI becomes embedded in everything from homework tools to entertainment feeds, protecting children from digital manipulation is getting harder. Parents and educators must go beyond “screen time” debates and help kids understand how AI-generated content is created, who makes it, and why. This piece argues that hands-on media creation—writing, recording, and experimenting with AI tools—is the best way to build awareness, skepticism, and resilience.

I’m a frequent user of AI tools like ChatGPT, so I pay close attention to the ongoing debates about what AI will do to us. As these tools become more varied and sophisticated, it’s clear they’ll have both positive and negative impacts across research, education, employment, and entertainment.

My view is that we must manage how we use AI. Like any rapidly developing technology, its effects will be both good and bad. Control will be complex, and there will be bumps—some major, some minor—along the way. We can also expect surprises, along with calls for new laws, regulations, moratoriums, and outright bans.

While I’ve been extremely pleased with how I’ve used AI tools personally and professionally, I’m no cheerleader for unfettered access. Rules and guidelines are necessary. Their effectiveness, though, will depend on how smart we are—and how willing we are to resist deception and political posturing.

There’s no doubt that jobs will be affected. Some workers will benefit; others will be hurt—blue-collar and white-collar alike. My attitude is simple: get on the train or get out of the way if you want to maximize benefits and minimize harm to yourself and your family.

These thoughts aren’t unique. Anyone whose career has involved developing, managing, or using technology has likely had them. What concerns me more is our responsibility to those who aren’t as immersed in technology—especially our children.

How do we want AI to affect them? And can we actually control that influence, given how deeply AI is being woven into every part of their world?

These are difficult questions. Consider how complicated the issue of “At what age should children get their own smartphones?” already is. Now add to that the limited control parents have over the systems and content kids can access—and how easily AI tools can create or alter the content they experience.

It won’t be possible to “screen out” AI-generated material, nor should we try to block legitimate educational uses of AI. The real issue is preventing technology-enabled harm. That means everyone—parents, teachers, and children—must understand what they’re exposed to and how AI shapes that content.

Ultimately, kids and parents alike need to be educated media consumers. They should understand how media are created, by whom, and for what purpose. The best way to teach that is through hands-on experience: writing, recording audio, and creating videos and animations—all media that AI can now help produce or manipulate.

Children need to recognize that media can promote not just facts, but opinions and biases as well. Hands-on experience is the best way to help them learn how hard it can be to tell what’s real, what’s fake, and what’s being distorted.

Copyright © 2025 by Dennis D. McDonald

Related Topics

According to a recent Science news report, Journal Publishers crack down on research from open health data sets, some research journal publishers are looking twice at suspect research based on public health data sets such as the CDC's National Health And Nutrition Examination Survey (NHANES).

The problem: some researchers are submitting papers containing potentially spurious correlations that, unless confirmed by additional analysis, can communicate potentially misleading findings related to public health.

Solutions to the problem, which, according to the Science article has resulted in multiple journals already having rejected literally thousands of submitted papers-– include:

• Requirements for additional analysis.

• Inclusion of results from different data sets.

• Compliance with checklists and extra steps that must be undertaken by researchers prior to submission.

• Outright bans on acceptance of papers from selected open data sets.

While the report also states that some "paper mills" are already finding ways to sidestep some of these control methods, the bottom line is that journal editors and reviewers will now have extra work to do as long as a "publish or perish" ethos is imposed on researchers.

My guess is that many editors have already recruited AI tools to aid in the review process to help identify and “weed out” questionable submissions. That “open science” has both positive and negative implications is nothing new. The irony of this in our AI-soaked world leads me to wonder if we are heading towards our own Butlerian Jihad?

Copyright 2025 by Dennis D. McDonald