Will AI make us worse people?

Multiple writers have highlighted concerns that AI is contributing to what we all might agree is bad behavior. In this blog I highlight several of the areas of concern regarding the negative impact of AI on our behavior. Specifically, I will consider whether AI is likely to contribute to our becoming: more dishonest; less open-minded and willing to consider other points of view; and less likely to think critically due to a decline in reading. My use of the word “concern” is a deliberate choice, suggesting the possibility of real problems arising from AI, rather than definitively declaring their existence. In addition, AI is far from the only culprit that is likely contributing to these ills. In addition, the concerns that I have raised, and which have been noted by multiple writers and researchers, does not preclude potential meaningful benefits for AI. In a subsequent blog I will consider the potential benefits of AI in more detail. 

One of the most blatant examples of a way in which AI makes us worse involves cheating and plagiarism. A recent article in the Atlantic magazine (1) bemoaned the end of the honor code at Princeton. Specifically, the article details how increased concerns about students cheating on exams resulted in a return to proctoring exams. Princeton is obviously an elite university and I can imagine some arguing that the sense of entitlement of Princeton students might be a factor. If Princeton was the only example of this I might entertain this argument, but that is far from the case. A recent New York Times article (2) notes concerns about the rise in cheating and details how there is a growing sophistication to AI based cheating, including programs such as “Grubby AI” that apparently writes papers in ways that make them look more like a student wrote them, including making minor errors. The article noted multiple programs that claim to be useful in evading AI detection programs as well as noting the rise of influences and promoters of these programs. However, other researchers have cautioned about making assumptions about AI use and cheating, suggesting there is much complexity in how AI is used, and variation among students in terms of who uses AI more and who is more likely to use AI to cheat (3). One finding that stood out from the Science (3) study, that was highlighted in the Cornell Chronicle, was that it appears that more frequent users of AI are more likely to cheat (4). Moreover, the authors of the Cornell Chronicle article suggest that as AI use grows the risk of more cheating will also increase. 

Sadly, cheating is not limited to college and high school students. Recent controversies have erupted regarding a number of books and authors, some well respected until scandals about their use of AI have erupted (5). However, accusations about the use of AI are controversial because it is difficult to prove AI was used. This is best reflected in the controversy surrounding the Commonwealth prize for the best short story in which the author was accused of using AI, and then the accusations were withdrawn (6). Again, as is the case with student cheating, the data is not sufficiently solid to confirm widespread cheating and plagiarism, but there is clearly a strong suggestion that these risks are on the rise, and may continue to grow. 

AI may also result in us becoming more certain of our views and less open to considering alternative points of view.There appear to be a number of ways that this can occur. First, AI (specifically Chatbots) has been repeatedly identified as sycophantic (7), as validating our beliefs and points of view even when we are misguided, uninformed, or even factually wrong. The most tragic examples appear in news reports of individuals who have had problematic and delusional beliefs reinforced by Chatbots (7). In addition, there is evidence suggesting that Chatbots can have a more subtle, but nonetheless pernicious influence on our willingness/capacity to be empathetic and to resolve conflicts. This concern has been repeatedly raised about social media, and how the algorithms that determine what content we receive, highlight more extreme and provocative materials. A recent study in the journal, Science (8), suggests that Chatbots may have a similar impact. This study details how sycophancy, the tendency to validate and affirm what is said regardless of its validity or morality/ethicalness is a feature of AI (specifically large language models and their chatbots), and has multiple detrimental effects: reductions in the willingness to take responsibility for one’s actions, decreases in openness to repairing or fixing interpersonal conflicts, and a tendency to reinforce the belief that one is in the right regardless of the course of action proposed or reported. Moreover, this study showed that users preferred the sycophantic responses over human responses that were more challenging and critical. While there clearly is a need for more studies there is other preliminary research that supports this finding. Recent research conducted at Stanford University suggests that therapy focused Chatbots potentially have a negative impact on mental health (9). 

One more concern is the possibility that AI is making us intellectually lazy. While I cannot blame AI for the decline in reading because it clearly began before the advent of AI, I am concerned that AI is accelerating that decline. Recently, I encountered several articles that have documented a decline in reading, for all age groups (10a, 10b, 10c). The implications of this decline are elucidated in a thought provoking, and albeit depressing article, by Rose Horowitch, in a recent edition of the Atlantic magazine (10b). Horowitch suggests that the decline in reading has broad implications for how we understand and approach and understand the world we live in. Horowitch cites Maryanne Wolf, a cognitive neuroscientist at UCLA, who, per Horowitch “argues that people are losing the ability to think deeply about what they write. That doesn’t mean they are forgetting how to decode individual words. Rather, they are losing the higher-order abilities of comprehension and synthesis.” Specifically, Horowitch, and many others, are concerned that our capacity to think deeply, thoughtfully analyze problems and draw inferences and conclusions, are all declining. Moreover, she argues that this problem is only going to increase because of the declines in reading by children, teens and young adults. 

Is there hope? While my bias is clearly showing: I am worried that AI will make us worse, it is possible that this can be avoided and that the benefits and potential benefits of AI will outweigh any negative issues. In my next blog, “Do not despair, AI has real potential” I address ways in which AI may be a boon. 

References:

1. The end of the honor code at Princeton and the rising concerns about cheating. The Atlantic article details how Princeton recognized that AI was leading to increased cheating and not only began to proctor exams by reducing the use of takehome exams substantially. https://www.theatlantic.com/ideas/2026/05/princeton-ai-honor-code/687144/
2. NYT article on rise in programs designed to facilitate cheating, specifically on programs that can write essays and papers in ways that are not likely to be detected as written by AI. https://www.nytimes.com/2026/06/18/us/ai-apps-students-cheat.html
3. See the article, “Generative AI use and misuse call for assessment reform in higher education” in the Journal, Science: https://www.science.org/doi/10.1126/science.aec5115
4. The Cornell Chronicle article which discusses the Science study on generative AI use and cheating: https://news.cornell.edu/stories/2026/05/widespread-ai-misuse-means-higher-ed-must-rethink-assessment
5. Article detailing the accusations against the author of Shy Girl for using AI. https://www.cbc.ca/news/business/publishing-industry-ai-shy-girl-fallout-9.715529. There was also a NYT article that discusses this cancellation of Shy Girl. https://www.nytimes.com/2026/03/19/books/shy-girl-book-ai.htm. See also a list of books, compiled by the website, GoodReads, which have been identified as likely being AI generated. https://www.goodreads.com/list/show/221372.Obviously_AI_generated
6.  See the article in the Atlantic that discusses this controversy and the challenges of determining whether a novel, short story or article is written by AI. https://www.theatlantic.com/technology/2026/07/commonwealth-prize-ai-writing-jamir-nazir/687806/
7. AI Sycophancy is fairly widely accepted (see a recent article in the Stanford report:  https://news.stanford.edu/stories/2026/03/ai-advice-sycophantic-models-research Moreover, there are clear risks associated with this as well as Chartbots have been implicated in increasing delusional beliefs as well suicidal ideation.  See a recent study regarding the risk of Chatbots increasing delusional beliefs https://arxiv.org/abs/2602.19141. 
8. The article in Science reviews a study which suggests AI use decreases conflict resolution and increases a tendency to maintain one’s belief even when presented with contradictory evidence. https://www.science.org/doi/10.1126/science.aec8352
9. Preliminary research suggesting that AI and more specifically Chatbots, can have a negative impact on mental health include the following: https://hai.stanford.edu/news/exploring-the-dangers-of-ai-in-mental-health-care, and there is also some evidence that when Chatbots act in quasi therapist roles that their responses can also be highly problematic: https://www.brown.edu/news/2025-10-21/ai-mental-health-ethics
10.  A number of studies and reports have documented a decline in reading for pleasure and reading in general.
    1. A recent study of reading for pleasure, “The decline in reading for pleasure over 20 years of the American Time Use Survey” found a 3% decline in reading for pleasure, per year, from 2003 to 2023.” More specifically, the authors found that “In a large nationally representative sample of the US population, the proportion who read for pleasure on an average day declined over the last 20 years, from highs of 28% in 2004 to lows of 16% in 2023. https://pmc.ncbi.nlm.nih.gov/articles/PMC12496190/
    2. Rose Horowitz, in a recent edition of the Atlantic magazine, discusses the decline in reading, citing a variety of studies and findings noting a decline in reading and in what could broadly be referred to as reading comprehension skills. She also notes that there is evidence to suggest that what is being read by today’s readers, on average, is less complex and challenging. She cites the declining reading scores for 4th and 8th graders, as an additional concern. Horowitz also suggests that we may be entering a post-literacy age in which we still read, but what we read is far shorter and much simpler, e.g., tweets, posts, emails and the like. 
    3. In a short piece in the Conversation, an online news site, the Amherst profession, Austin Sarat, laments the decline of reading in high school and college while also lamenting this decline. https://docs.google.com/document/d/1DxFspVEQWccKCSdGi778iR63Q-Sg9XTRDn067i43w3c/edit?tab=t.0