Four recent studies highlight items to consider as we embrace Generative AI tools.
- Microsoft and CMU: Found that confidence in generative AI for problem-solving tasks, reduced critical thinking. ("Specifically, higher confidence in GenAI is associated with less critical thinking, while higher self-confidence is associated with more critical thinking.")
- BBC: Found that "answers produced by the AI assistants contained significant inaccuracies and distorted content from the BBC." (51% of answers had significant issues,19% introduced factual errors,13% of the quotes were either altered or not present in the article cited.)?
- UC Irvine: Found that users consistently rated AI responses as more accurate than they were. ("experiments with multiple choice and short-answer questions reveal that users tend to overestimate the accuracy of LLM responses when provided with default explanations. Moreover, longer explanations increased user confidence, even when the extra length did not improve answer accuracy")
- National University of Singapore: Demonstrated mathematically that large language models have computational limits, meaning they will inevitably make mistakes. ("In this paper, we formalize the problem and show that it is impossible to eliminate hallucination in LLMs.")
I came across the studies via reporting from
404 Media
and
The Neuron - AI News
. To gain a quick perspective on the MSFT & CMU study, listen to the 404 Media podcast.?
Leadership and Organizational Coach | Sales-Focused Product Management Executive and Leader | Experienced Presenter & Facilitator | LGBTQ+ Inclusion Champion
1 周I think it depends on what type of tasks we've asked whichever AI chatbot for help. Writing emails, responses, or form letters does require some level of mental energy and time, but couldn't that time and energy be better spent on other work that needs more human attention? I remember when you and I were working together that you would tell me how you spent your evenings and even weekends clearing your inbox. If you could go back and have a GPT do that for you, and write relatively closely to your own voice, I'd imagine you would have loved to have that time back. Allie K. Miller, a friend and AI expert I follow coined the term "AI Bar": We know AI can find that perfect word, the cleanest sentence, the most compelling poetic visual, or whatever turn of phrase, that we start expecting that level of precision from ourselves. The AI Bar raises our standards, but it also makes imperfection somehow feel more painful. Perhaps as humans, we need to find that nuanced balance between leveraging the convenience AI provides for mundane tasks and recognizing the potential pitfalls of relying too heavily on its relative precision.
VP of Product Management ? AI, Fintech, SAAS, Search & Mobile ? Proficient in driving revenue growth, scaling & efficiency.
1 周AI is a tool, not a solution. The real problem isn’t just that AI can hallucinate—it’s that even critical thinkers struggle to detect when it does. Unlike a calculator, which is either right or wrong, AI operates in a gray zone where errors can be subtle, plausible, and hard to fact-check. AI’s "confidence" in responses definitely makes people overtrust it, even when it’s making things up. I'm not sure it's necessarily a failure of critical thinking—it’s just how human brains work. I've seen first hand and am aware of studies that show that the more detailed an AI response, the more accurate people think it is, even when it isn’t. I think the real challenge is how do we learn how to use AI while knowing it will always get some things wrong?