In the tech world one paper drove a huge number of headlines in the last week. Microsoft Research posted a paper looking at the impact of generative AI on critical thinking. It lead to some headline gems, including these:
- Your Brain On AI: ‘Atrophied And Unprepared’ – Forbes article.
- AI Is Making You Dumber, Microsoft Researchers Say – also Forbes.
- New Study Proves Generative AI Harms Critical Thinking – Huff Post (with the clickbaity/upworthy headline: Relying On AI All The Time Has This Surprising Consequence).
- Are we losing our critical thinking skills to AI? New Microsoft study raises red flags – ZDNet.
Actually, probably the main ‘red flag’ there is the line ‘new study proves…’ – it is incredibly rare for a single study to prove anything. There were dozens of other headlines, so I dug into the full paper (here). Makes for some interesting reading, although it is a survey-lead piece of research, based on self-reports. Obviously, given my day job, I am very happy with a good piece of survey-based research, and this has all the hallmarks of one. While the headlines don’t exactly line up with what the research paper says, it does provide another piece of evidence for what may of us have come to suspect over the last year or two: Heavy use of AI decreases people’s confidence in their own thinking ability. They are less likely to try to reason, and less successsful at doing so, after extensive use of generative AI.
Beyond the headlines it has produced, the most interesting part for me was the survey design, and the reminder of Bloom’s taxonomy1, which groups thinking and learning objectives into six types:
- Knowledge (recall of ideas),
- Comprehension (demonstrating understanding of ideas),
- Application (putting ideas into practice),
- Analysis (contrasting and relating ideas),
- Synthesis (combining ideas),
- Evaluation (judging ideas through criteria).
It makes for a useful framework to think about how we use generative AI, and context for how that impacts on our skills. Myself and many others have noticed “the blank page problem”, where AI excels at getting us from a blank page to a first potential draft, the more people use it for this, the harder and harder they seem to find it to break through the blank page barrier without resorting to AI. There are parallels here in the history of technology. Home appliances, the motor vehicle, and then the home office have almost completely eliminated natural exercise from many people’s daily lives – the impact of that is very evident in the health care system (Public Health England and NIH). Will there be an equivalent impact on cognitive ability? In the future will we need ‘mind gyms’ to have regular cognitive work-outs, to stop our brains becoming flabby? How about “Thought Club” – A place to get a mental work out and build your thinking skills? Have a mental stretch and do some heavy lifting?
The impact of generative AI is most likely to be a decline in thinking. However, there is an opportunity to use the outputs to take our thinking to another level, standing on the shoulders of giants (or more-likely on the shoulders of the average Jane/Joe). That takes significant effort though, and it isn’t the default. To stay good at thinking in the long-term, we need to get better at understanding thinking in the short-term.
[AI generated post image used to prove the point of Better Images Of AI].
Foot notes:
- Benjamin S Bloom, Max D Engelhart, Edward J Furst, Walquer H Hill, David R Krathwohl, et al. 1956. Taxonomy of educational objectives: the classification of educational goals: handbook I: cognitive domain. Technical Report. New York, US: D. Mckay.
William Huitt. 2011. Bloom et al.’s taxonomy of the cognitive domain. Educational psychology interactive 22 (2011), 1–4. ↩︎
