Arguably, the most important outcome of a university education is the expansion of a student’s capacity for thinking critically about ways to solve problems. Outsourcing that thought process to any form of artificial intelligence (AI) not only precludes the growth of critical thinking skills, but also erodes your existing skills.
GPT5 and Claude are large-language models (LLMs) designed to predict responses in a conversation without regard to the veracity of the information that they are providing, or the reliability of the sources of information. Given limited information, LLMs will fabricate information to fill in the gaps, without informing the user that it is doing so. The tendency to fabricate cannot be altered by user prompts demanding veracity. The only programmatic guiderails that exist to promote veracity are designed to protect the product’s reputation, not to protect the user. The confidence with which information is presented to the user by LLMs is based upon the confidence of the algorithm that it has predicted the next response, not by any metric of the veracity of the statement.
The processes that make LLMs effective at predicting responses are in direct opposition to the fundamental principles of critical thinking. LLMs will not evaluate what additional information is required for a reliable response, which is why they do not ask clarifying questions before providing a response. LLMs start a new reasoning chain for every new question, which can result in fundamentally different answers to the same question asked in different sessions. LLMs often will not re-evaluate a reasoning chain when presented with a contradiction because doing so is detrimental to its goal of predicting responses; it will instead justify that chain of reasoning.
For these reasons:
LLMs are a terrible substitute for a search engine.
LLMs are a terrible choice for problem-solving.
Apart from the issue of claiming someone else’s work as your own, using LLMs for assignments creates the risk of facing consequences for academic dishonesty because of their tendency to fabricate information.
The use of LLMs or any other form of AI for any gradeable material in this course will result in a grade of zero for that material.
Send comments, suggestions, and corrections to: Derek Zelmer