ai

An Important Note for My Students Regarding the Use of ChatGPT

Douglas McManaman

I have listened to lots of people over the years – most of them professionally and some just because we were together. I don’t know of one person among those many who, under the right conditions, didn’t have interesting and important things to say. Colleagues sometimes ask me whether I ever get bored listening to people. Yes: under one circumstance. If people are not saying what they really think, when they are chronically ducking and censoring or trying to impress or placate, I am bored. But if people are thinking for themselves about things that really matter to them, I am fascinated. You can tell when a person has just moved from let-me-please-you thinking back to their own mind – they go from soporific to scintillating just like that. I enjoy those moments hugely. Beneath the fear of being punished for thinking for themselves, most people have ideas that matter, ideas that would make a difference if they could be developed fully. People, regardless of their position or status, can think of things that move discussions to whole new levels of sparkle and resolution. Individuals you would never suspect of being interesting have absorbing stories to tell and disturbing insights that would humble even the most long-winded of us right out of our self-importance and rush. If the conditions are right, the huge intelligence of the human being surfaces. Ideas seem to come from nowhere and sometimes stun us. 

Time to Think: Listening to Ignite the Human Mind by Nancy Kline

Recently I came across a journal assignment that had an opening sentence that was word for word identical to that of another student. I mentioned this to my daughter and she just smiled. She said: ChatGPT. I asked her to explain to me how this works, so she took one of my journal assignments that my students are required to complete and showed me how it is done, by actually inputting the url links to the articles I had assigned and the questions I wanted them to cover. Lo and behold, a very well written journal assignment that summarizes an article that I wrote, as well as an article from a recent pope, and which incorporates answers to the questions I assigned. The AI product even included a line that began with: “I learned that…”, a line that I see very often from my students.

Now, I have no problem with students using AI in order to procure basic information–I will often use it myself to summarize a book that I am interested in purchasing, to see if I want to pursue the book further. The problem with using ChatGPT to do an assignment for you, however, is fundamentally a moral problem, and a serious one at that. I’d like to explain this, because for some reason, I had a difficult time getting my daughter to understand this, which made me wonder whether some of my students would also have the same difficulty. 

So, allow me to make two points. The first is that using ChatGPT to complete an assignment for you is really no different than having someone do your homework, essay, or journal for you. In short, it is a matter of cheating, which in turn is a matter of lying. This brings me to my next point: you and I determine our moral identity, our character, by the moral choices that we make. Moral character is not the same as one’s personality. Character describes the kind of person that you are, the kind of person that you have made yourself to be by the moral choices you have made and continue to make in your life. So, a person can have a very nice personality, but bad moral character–he or she might be a liar, which is a person who cannot be trusted. And I become a liar by choosing to lie, or a thief by choosing to take something that does not belong to me, or a killer by actually killing someone, or an adulterer by choosing to commit adultery, etc. 

What frightens me about students who use ChatGPT to complete a journal assignment that involves reading an article and/or watching a video is that they are becoming liars, persons who cannot be trusted. Now, this is a problem because every moral choice we make comes back to us in the end, either to bless us or curse us. It was always a challenge to get my young grade 9 students to understand the harm they do to themselves when they lie–sure, they often get away with it, but the deficiency in their character which their choice brought about stays with them until they make a choice inconsistent with their previous lie–which is what it means to repent, that is, to have a change of heart. But it is disconcerting to think that some adults, not all, need to have this pointed out to them. One’s character is one’s “heart” (the will), and so a good character is a good heart. I’ve always said to my students that character is everything. Our eternal destiny is determined by our character, and the more I go on in life, the more I see that the problems of this world for the most part boil down to matters of character. There is so much suffering in this life as a result of deficient moral character. What is at the root of failed marriages? Very often it’s a matter of bad character, i.e, someone’s infidelity, or impatience, or taking oneself too seriously, selfishness, etc. What is at the root of traffic jams? Often it is a matter of construction, but too often it is a matter of a serious vehicle accident up ahead, which in turn was the result of impatience, which is a moral vice–someone was in too much of a hurry and was driving recklessly. No need to spell out the root causes of war, poverty and hunger. Everything we do comes back to us in the end. Eastern religions referred to this as karma, which is a teaching that expresses a universal experience.

The purpose of an assignment, in my case a religion assignment, is to have my students acquire the knowledge and intellectual disposition that they will need in the classroom, in order to create the conditions in their own classrooms that will allow students to question further and build on what we give them, which often leads to new avenues to explore and new insights. All this gets thwarted when we choose to use AI tools to do the work for us.



Some Thoughts on Sanity, Theology, and Change

Copyright © 2020-2026 by Douglas P. McManaman
All Rights Reserved (revised in 2026)

Deacon Douglas McManaman

We speak of psychosis as a loss of contact with reality, either permanent or temporary; psychotic episodes, for instance, are temporary breaks with reality. There is a sense, however, that being out of touch with “reality” is a matter of degree. I contend that the more we come to understand the inductive nature of knowledge acquisition and its implications, we should begin to see that we are always, to some degree at least, out of touch with reality. All knowledge begins in sensation, as Aristotle maintained (nothing is in the intellect that is not first in the senses); in other words, our knowledge has empirical origins; this means it begins with evidence and proceeds towards the most coherent and consistent explanation of the evidence. Most importantly, however, although our grasp of the real expands continually—or should—, the process of expansion is—or should be—accompanied by an awareness of an ever-decreasing circle in the midst of which we find ourselves, and at the edge of which is a vast and expanding penumbra of obscurity. Leaving aside early or first-episode psychosis (FEP), the genuinely insane are typically unaware that they have lost contact with the real, but the most sane among us have the greatest awareness that in the final analysis, the reality they are in touch with is so much larger than is their current grasp of it—increasingly so—, and they have the greatest awareness that their current worldview, which cannot exceed the limited information they possess about the real, is in large measure the product of what they believe the world to be through the lens of that limited set of information. It is all too easy to confuse our worldview with the world, which we always know only deficiently.

The benefit of coming to a deeper appreciation of statistical reasoning and Bayesian inference [1] is that one begins to realize both how risky our intuitive and formal statistical inferences are, and how precarious are those estimates that are the product of Bayesian inference. Statisticians and research scientists tend to have a deeper appreciation of the risky nature of their (and our) current convictions. Their subject matter is very often data that is too large for us to manage with great precision, such as a population mean, which we can only estimate—along with its standard deviation—, on the basis of a sample. The estimate, however, is typically an interval, and the wider the interval, the greater our confidence; the lesser our confidence level, the narrower and more precise that interval becomes. In other words, the more precise and thus more useful our estimates, the greater their vulnerability to error (i.e., we estimate the house for sale down the road to be between $850,000 – $900,000 vs. between $100 and 6 million. The latter is more certain, but less useful). Perhaps this is why good scientists tend not to speak with a rhetoric of high confidence, especially when our truth claims bear upon matters of precision. Only when matters become more general—and perhaps less useful—is greater confidence warranted. 

Moreover, Bayesian inference should bring us to a greater awareness of the role that experience plays in knowledge acquisition.[2] The probability of a hypothesis given the evidence [p(H|E)] leaves us with a space of uncertainty (i.e., 49% or 70%), and such inference depends upon a knowledge of base rates (for example, 48.2% of all Americans age 15+ are married, while 51.8% are not) and likelihoods (i.e., the probability that a couple has children given that they are married is 90%). However, when we estimate the probability of a belief given certain pieces of evidence, we tend to ignore base rates (prior ratios), and so our estimates that are the product of intuitive reasoning are often seriously mistaken—hence, we ought not to trust our intuitive probability estimates. The most important implication of Bayesian inference, however, is the effect that experience has on our posterior probabilities: new information changes them. And so, once again, we are reminded not to be too confident in what we claim to “know”—the very fact that new information that affects our base rates demands that we continually update our estimates, not to mention recognize them as estimates in the first place, and not “knowledge” per se.[3]

Our day to day reasoning, however, is not fundamentally mathematical, that is, we do not typically perform mathematical calculations on the basis of prior probabilities; rather, we reason on the basis of plausible data, not probabilities, and the reasoning is not calculative, but comparative. Sometimes what is improbable, i.e., that Jack was hit by a city bus, is moderately plausible given the plausibility indexes of our current data (witness statements, or a statement from the victim, or other sources, etc.), and often a number of competing estimates are equally plausible.[4]

In terms of plausible reasoning, all we ever have at any one time are limited sets of data formulated in propositions having a degree of plausibility, either minimal, moderate, high, etc., that is, a less than certain character. The entire set of data at our disposal is typically overabundant and inconsistent. Indeed, there are many propositions in our data set the truth of which we can be certain and from which we can deduce a great deal, rendering explicit what was previously implicit.[5] There are, however, a myriad of theses that are less than certain. The task of sound reasoning is to bring maximal consistency to this set of data.

What is particularly interesting to note is that bringing maximal consistency does not guarantee that in the end we possess the truth. What we have at best is the most plausible estimate given the information currently available. New information very often alters the consistency of our plausibilistically favored subsets of data bearing upon specific matters, with the result that a new estimate is in order. This is why there is a great deal of “mind-changing” in the sciences—we just don’t know whether or not we have enough information at any one time to resolve a particular question with complete certitude. It seems, in fact, that we are always information deficient.

And so, once again, the worldview that results from our current set of information is an ever changing one, that is, an evolving worldview. It has always been a deficient worldview, because the information on the basis of which it is established at any given time is deficient. Even the little that we have at our disposal is a product of interpretation, and our interpretation is once again made up of risky inferences. As Feynman says of science, it is an ever-expanding frontier of ignorance. Similarly, our day to day knowing is precisely an ever expanding frontier of ignorance: the more we come to know about the world we live in, the more we should realize just how much more we did not know than we previously thought there was to know. With every new discovery comes a manifold of new questions, and new questions open up new and unexplored avenues that, when explored, provide new information that very often upsets the consistency of what we thought was a well-established conceptual framework, causing us to adjust our estimates by discarding data inconsistent with more plausible data in order to establish a different and plausibilistically favored subset of data from which a better and more accurate worldview may arise. Moreover, new information may inadvertently strengthen a position we’ve held for a time; however, a new and maximally plausible estimate is no guarantee that we are any closer to the truth—a previous but now plausibilistically less favored estimate may in fact be true, and time may reveal that. In other words, the most current estimate is not necessarily closer to the truth. That is why learning is very often an oscillating process. If a position or estimate is true, it is not necessarily the case that newer information will corroborate it; we may be taken further away from the truth, only to return to it at a later date. Progress, in other words, is not necessarily unidirectional.

What this implies is that we are always, in a manner of speaking, out of touch with reality, for reality is so much larger, inconceivably larger, than our current grasp of it, and the frontier of our ignorance is ever expanding. And although we are always relatively out of touch with the real, at least we can know that we are always relatively out of touch with it. That, I contend, is what distinguishes the sane from the insane—the insane are out of touch and have no awareness of the fact. I dare say, however, that most people believe their grasp of reality to be far more comprehensive that it can possibly be, for many speak with a rhetoric of certainty that assumes a knowledge that is just not humanly possible on a large number of issues, given the little time invested in those matters. What I am suggesting is that most people have a greater resemblance to the insane than they do to the genuinely sane; the former tend to resist this never-ending learning process that requires adjusting our estimates in the light of new information. The intellectual, for example, who works exclusively in the realm of ideas, who has little interest in testing those ideas before they are imposed on a society, has a greater resemblance to the insane than the sane, which is likely why intellectuals who succeed in having their untested albeit interesting ideas implemented on a wider social scale usually end up costing the taxpayer a great deal. 

One irony in all of this is that a great deal of disordered confidence and resistance of the learning process is found within that discipline whose object is the mystery par excellence, namely the unutterable mystery of God. Many who are fond of theology fail to appreciate just how much the logic of the scientific method is involved in this more general science. Moral philosophy does not escape this logic, nor is this logic foreign to biblical exegesis and the study of Scripture, and thus by extension, moral theology or any other branch of sacred theology. Moral reasoning follows much the same law of complementarity that we encounter in statistics: the more universal or general the discourse, the greater the certainty, but as we move to greater precision, vulnerability to error increases. A fine example of this is Germain Grisez’s Difficult Moral Questions (Franciscan Press, 1997). That volume was the product of years of thinking about principles and their application to specific moral problems that have arisen as a result of new circumstances. On a number of occasions, I had the privilege of observing Joseph Boyle’s uncertainty as he pondered on the edge of the frontiers of a difficult moral problem. He refused to overstate his case, and he was all too aware that he may not have in his possession enough rational data necessary to satisfactorily work out the problem which preoccupied him at the time; moreover, these analytical moral philosophers (Grisez, Boyle, Finnis, etc.) have, over the years, changed their position on a number of important issues, thanks to more thought, dialogue, and discussion. Most especially, we see the same inductive/investigative process in the area of biblical studies/exegesis. With new historical data, what was once thought to be the case is now relegated to a lower level of plausibility while a more plausible hypothesis takes top spot. 

Canon lawyers working on marriage tribunals, for example, judging cases on a team of three, will testify that some cases are easy while others are very difficult; the latter are often resolved with a 2:1 ratio (the one outvoted has to humbly accept the majority decision, but after reviewing the reasons given will often see what was not noticed earlier). Those on the outside, unfamiliar with this process–and it is a process–, tend to have a difficult time appreciating the subtleties of these matters. Unlike judges who regularly work on such cases, most people have not encountered such intricate and murky situations, permeated as they are with uncertainty. A view from the “inside”–whether the subject matter is politics or law, etc.–is very different from a view from the “outside”, and many on the outside are too emotionally vested to acknowledge their own deficiency of information and will proceed to dogmatically spout off on all sorts of issues they know very little about.

Pastoral approaches to spiritual direction magnify this logic even further. A good pastor of souls must be able to pick up subtle clues in the words, gestures, and reactions of the directee, clues on the basis of which one may rapidly inference to information needed to uncover the best way to communicate important principles and insights that cannot be effectively imparted in the same way to everyone. A pastor of souls, like a good teacher, is one who is capable of detecting clues that give evidence of conditions within a person that render him/her temporarily incapable of understanding certain things (as well as conditions that make possible a certain understanding). Moreover, there is a distinction between a pastor of souls and a moral theologian. It is certainly possible for a person to be both, but a theologian without a good pastoral sensibility, that is, without a mind for contingent factors and other clues and who perhaps loves moral problems more than people, is not someone who should be providing spiritual direction. To be avoided are the two extremes of the easy going nonchalant who confuses a pastoral sense with moral permissiveness on the one hand, and the hard-nosed dogmatist who has no sense of the complexities of the human person on the other.

There’s no warrant for dogmatism here; what appears to be the “truth” at one time is often eventually discovered to be a rather deficient position or a position in need of further distinction. In the end, what this suggests is the need for a spirit of greater humility; it suggests the need for constant dialogue and a listening posture. But this is precisely the posture lacking in a large sector of our society, including our Church, that is, among the passionately conservative or traditional, among many of the clergy (both “liberal” and “conservative”), as well as the university environment, among professors of certain non-scientific disciplines, etc. What makes these epistemic matters more difficult is the fact that character, psychology, and mood play a significant role in knowledge acquisition. Character plays a fundamental role in our ability to make moral distinctions, among other things—people will not see what it is they are unwilling to see or are not emotionally ready to see. Character is a more permanent epistemic condition, while mood is temporary. Both, however, can beget blind spots.

End Notes

1. Bayesian inference seeks to estimate the probability of a belief or hypothesis given certain pieces of evidence. Hypothesis testing, on the other hand, seeks to determine the probability of evidence given a particular belief or hypothesis (a null or alternative hypothesis).

2. The formula for Bayes Theorem is: p(H|E) = p(H)p(E|H)/p(H)p(E|H) + p(~H)p(E|~H)

3. We typically confuse the p(E|H) with p(H|E). For example, over the years I have found that school administrators often assume that since the likelihood that a good teacher interviews well is over 90%, it follows that this or that person just interviewed is a good teacher (90% probability), since she interviewed very well. This conclusion, however, is invalid. There is a real distinction between 1) the probability that a person interviews well given that she is a good teacher [p(E|H)], and 2) the probability that this person is a good teacher given that she interviews well [p(H|E). For the sake of argument, let it be the case that 90% of good teachers interview well–that’s not an unreasonable assumption. Furthermore, with some experience in education, it soon becomes evident that the majority of a typical staff of teachers are not great teachers–great teachers are usually in the minority, and administrators desperately want to hang on to such people when they discover them (let us say 20% are hard-working, self-motivated, reliable, positive, love their subject matter and their students, and are not in it merely for the perks, etc.). And let us estimate that the likelihood that a “not so great” teacher will interview well is 40%. Given these numbers, the probability that this person is a good teacher given that he/she interviewed very well is only 36% (0.2 x 0.9/0.2 x 0.9 + 0.8 x 0.4 = 0.36). Hence, the reason administrators, much to their dismay, continue to hire the wrong people. Bayesian inference requires that we pay attention to how our prior probabilities change with the addition of new evidence. 

Consider as well how judgment of character might look from a Bayesian point of view. I see a person for a relatively short period of time. I am not aware of this at the time, but he’s going through chemotherapy treatments, which can make it much easier for a person to behave in a way that is relatively uncharacteristic. But during this relatively small period of time, he gives evidence of undesirable character. A person of good character might give evidence to the contrary about 5% of the time overall, but within a relatively limited period of time, he might give evidence to the contrary about 50% of that time period. After a while, it might average out to about 5% (much like scoring birdies for the first four holes in a game of golf, only to average out to 102 by the end of 18 holes). More time and experience allow us to change our prior ratios, that is, our base rates. A 1:1 ratio at the start of an investigation may become a 1:20 ratio by the end. Given a likelihood ratio 90:5 (a 90% likelihood that a person of bad character will give evidence consistent with it, and a 5% likelihood that a person of good character will give evidence to the contrary), a judgment that this person is of undesirable character can go from a 95% probability (highly probable) to 47% (i.e., the inference is probably wrong). The problem is that we typically focus our attention exclusively on the likelihoods [p(E|H)] and we neglect the incompleteness of our base rate information, that is, possible background knowledge that can change our posterior probabilities. 

4. It is not easy to explain the difference between Bayesian reasoning and plausibility reasoning. The former is quantitative and calculative, while plausibility is qualitative and comparative. Nicholas Rescher writes: “On the basis of logic and probability theory one cannot tell what may reasonably be accepted in the face of imperfect, indeed conflicting data. By contrast, the mechanisms of plausibility theory are designed to provide a basis on which it becomes possible to effect a transition of this nature–a move from the reliability of sources to the plausibility of their declarations. In providing a tool for handling cognitive dissonance, plausibility theory affords a reasonable basis for discriminating between the inferences which can and cannot be drawn from the inconsistent data-base yielded by the conflicting reports of imperfect sources. Accordingly, plausibility is intended to reflect an index of what reasonable people would–and should–agree on, given the relevant information.” Plausible Reasoning: An Introduction to the Theory and Practice of Plausibilistic Inference. The Netherlands: Van Gorcum, 1976, pp. 4-5.

5. This is particularly the case when it comes to probabilities and statistics. At the very least we can say that if our numbers are correct–and we cannot always be sure–, then we are certain that the interval is between # (lower limit) and # (upper limit). Moreover, I have argued elsewhere that on the most important matters, certainty is much easier to achieve. For example, the fundamental principles of the natural moral law (intelligible human goods) are naturally known, as well as the most general precepts of natural law. Indeed, they are imperfectly understood and inconsistently applied by most people, and of course much better understood by analytical moral thinkers who offer tentative estimates on the most difficult moral matters. And Leibniz has shown that the most important knowledge of all, namely the knowledge that God exists, is so simple that it is easily overlooked by most people: “If the necessary being is possible, then the necessary being exists” (because the necessary being cannot not exist, otherwise it is not the necessary being, but a contingent being). On matters somewhat less important, however, we are almost always information deficient. This epistemic state of affairs demands a posture of constant readiness to listen, to dialogue, to be corrected, one that is certainly not very widespread today, even in environments in which this openness to learning is reasonably expected to abound, namely the university environment.