Fact vs. Fiction: Information Integrity & AI

[Note: these questions were written in the summer of 2026. It seems that every aspect of AI’s development, usage, and impact is evolving at a rapid pace. Take that into account and use relevant, recent information in formulating your answers and arguments.]

1. Many of the speakers in this video emphasize the importance of accurate information for rational decision making in every realm of our lives—politics, finance, medicine, and so forth. Do you agree with this? Why or why not?
2. Given that our large language models (LLMs) are trained not to output facts but to output the next word in their training data, are you surprised that AIs hallucinate? Was this something you were aware of before watching the video? How do you feel about this? Have you ever been misled by such a hallucination? Describe.
3. As LLMs become more powerful and sophisticated, they are tending to hallucinate less. However, the hallucinations that highly advanced AI models turn out are much more difficult to spot. Are you comforted by the fact that many AI models seem to be getting a little more competent in regard to hallucinations? What do you think about the increased difficulty spotting hallucinations in these more powerful LLMs? Should AI companies be held responsible for the hallucinations that their models produce? Or is it solely the responsibility of the user to vet AI’s responses? Explain.
4. Certainly, all consumers of information these days must be more skeptical of what they see and hear online and elsewhere. They must think more critically than ever before. But is that enough? As AI-generated propaganda, lies, and “false facts” become more and more credible and increasingly widespread, do human beings need help from government regulators who might prevent or at least punish those who are behind this trend? What might that regulation look like?
5. Should creation of deep fakes and other forms of misleading or false information be protected by the First Amendment’s free speech protections? Why or why not?
6. One speaker talks about the challenges in training AI models that crawl the web to learn and thus accumulate all kinds of falsehoods along with facts. Do you think an LLM should definitively say that the earth is round? Why or why not? Do you think it should definitively say that Joe Biden won the 2020 presidential election? Why or why not? Support your argument with evidence.
7. In one study, researchers conjured up a fake condition that they called “bixonimania” and then seeded the internet with studies they had faked which included many hints as to their lack of genuineness—fake universities, fake cities, and even an article that stated: “This entire paper is made up.” Yet, when the researchers made AI searches as a person seeking health advice might do, many chatbots cited the study as if it were real. How do you feel about this? What do you think about the risks this could generate? How would you like policymakers to react?
8. Some companies are trying to design health-specific chatbots hoping to put in special features to guard against the types of errors noted in the previous question. Should they be allowed to sell these products before proving to regulators that the problems have been fixed? What is your opinion? Give reasons.
9. Nobel Prize-winning economist Joseph Stiglitz and a colleague modeled what they think is happening to cause all this manipulation of information and disregard for the truth. For a long-time, most information came to the public through respectable media with real journalists whose product was kept available (barely) by advertising revenue. Reporters spent actual time and money to check sources and learn the truth. When social media came along, a lot of money could be made by absolutely anyone who could produce “news” for public consumption with little and even no information just by making something up and putting it on the web. So-called citizen journalists and influencers could drive eyeballs to their digital platforms without doing any research at all. They could simply make things up or repeat rumors and conspiracy theories they had heard. The more provocative, the more outrageous, the more polarizing the information, the more money they would make. There is no economic incentive for these folks to spend money fact-checking or trying to get anything right. And AI enables them to send out vast doses of this “AI slop” with little or no extra cost. Does this model seem to you to be representative of what we are seeing today with AI? Why or why not? Support your argument with examples.
10. If the model described in the previous question is correct, or anywhere near correct, what can we possibly do about it? Are their risks associated with not doing anything about it? If so, what are they?
11. The Joint Research Center of the European Union has emphasized that AI has the potential to truly help users find and debunk false information on the internet. Do you find this comforting. Why or why not in light of the research you have done? Be specific.

We are currently working on a case study related to AI and information integrity. Be the first to know when it’s released by subscribing to our newsletter.

Click here to find our other case studies related to artificial intelligence.

AI Ethics Companion Handbook

This e-book compiles all the resources on AI ethics available on the Ethics Unwrapped website. It will continue to be updated as new materials are published.

A recent resource from Ethics Unwrapped is a book, Behavioral Ethics in Practice: Why We Sometimes Make the Wrong Decisions, written by Cara Biasucci and Robert Prentice. This accessible book is amply footnoted with behavioral ethics studies and associated research. It also includes suggestions at the end of each chapter for related Ethics Unwrapped videos and case studies. Some instructors use this resource to educate themselves, while others use it in lieu of (or in addition to) a textbook.

The most recent article written by Cara Biasucci and Robert Prentice describes the basics of behavioral ethics and introduces Ethics Unwrapped videos and supporting materials along with teaching examples. It also includes data on the efficacy of Ethics Unwrapped for improving ethics pedagogy across disciplines. Published in Journal of Business Law and Ethics Pedagogy (Vol. 1, August 2018), it can be downloaded here: “Teaching Behavioral Ethics (Using “Ethics Unwrapped” Videos and Educational Materials).”

Cara Biasucci also wrote a chapter on integrating Ethics Unwrapped in higher education, which can be found in the latest edition of Teaching Ethics: Instructional Models, Methods and Modalities for University Studies. The chapter includes examples of how Ethics Unwrapped is used at various universities.

An article written by Ethics Unwrapped authors Minette Drumwright, Robert Prentice, and Cara Biasucci introduce key concepts in behavioral ethics and approaches to effective ethics instruction—including sample classroom assignments. Published in the Decision Sciences Journal of Innovative Education, it can be downloaded here: “Behavioral Ethics and Teaching Ethical Decision Making.”

A detailed article written by Robert Prentice, with extensive resources for teaching behavioral ethics, was published in Journal of Legal Studies Education and can be downloaded here: “Teaching Behavioral Ethics.”

Another article by Robert Prentice, discussing how behavioral ethics can improve the ethicality of human decision-making, was published in the Notre Dame Journal of Law, Ethics & Public Policy. It can be downloaded here: “Behavioral Ethics: Can It Help Lawyers (And Others) Be their Best Selves?”

A dated (but still serviceable) introductory article about teaching behavioral ethics can be accessed through Google Scholar by searching: Prentice, Robert A. 2004. “Teaching Ethics, Heuristics, and Biases.” Journal of Business Ethics Education 1 (1): 57-74.