Reasoning With and About AI

AI systems are now part of everyday life — helping with homework, answering questions, writing text, generating images, and making recommendations. These systems can be extraordinarily useful. They can also be wrong in ways that are hard to detect precisely because they sound so authoritative. Knowing how to reason alongside AI — and about AI claims — is one of the most practical skills of this generation.

How AI Produces Its Answers

To evaluate AI reasoning, you first need to understand roughly how it works. A large language model (LLM) — the technology behind tools like ChatGPT and Claude — is trained on enormous collections of text. During training, the model learns statistical patterns: which words and ideas tend to appear together, how sentences are structured, and what kinds of responses tend to follow what kinds of questions. When you ask the model a question, it does not look up the answer in a database. It generates a response by predicting, step by step, which text most plausibly follows the input it received. This process is powerful for generating fluent, coherent text — but it is not the same as thinking through a problem from first principles. The model can be confidently wrong if the patterns in its training data were wrong, biased, or simply do not apply to your specific question.

Evaluating AI Reasoning: The Same Standards Apply

Every standard you have learned in this module for evaluating human arguments applies directly to AI-generated arguments. When an AI makes a claim, ask: what is the specific claim? What reasons does it offer? What evidence does it cite? Are there logical fallacies in its reasoning? Is there an obvious counterargument it is not addressing? The fact that AI generated the argument changes nothing about these standards. If anything, you should apply them more carefully, because AI output can be smoother and more confident-sounding than a typical human argument — which can make the flaws harder to spot on a quick read.

Pay particular attention to cited sources. AI language models sometimes fabricate citations — they generate author names, journal titles, and publication years that sound plausible but refer to papers that do not exist. Before using an AI-provided source in your own work, verify it independently. Search for it. Confirm the author, publication, and content are real. If you cannot find it, it may not exist.

Reasoning About Claims Made About AI

Just as important as evaluating AI output is evaluating claims that people make about AI. These claims appear constantly in news articles, company announcements, social media posts, and political debates. They range from the optimistic to the alarming, and many of them are poorly reasoned. Common patterns to watch for: exaggerated capability claims ('AI can now do X better than any human') that do not specify under what conditions the comparison was made. Catastrophic predictions ('AI will eliminate all jobs within five years') stated as certainties when they are speculative projections. Appeals to authority ('leading experts agree') without naming the experts or explaining their reasoning. And false dilemmas ('either we embrace AI fully or we fall behind') that ignore the nuanced middle ground most serious analysts actually occupy.