What research says about AI and thinking for yourself
In short: studies showing why it matters how we use AI, and how Monkey Score measures. As of November 2026.
The key points in three sentences
- AI can help or harm learning. What matters is whether you think along or simply accept answers.
- Several studies find that people who trust AI strongly or offload thinking often think less critically. Young people are particularly affected.
- AI with “guardrails”, meaning hints instead of finished solutions, largely prevents these downsides. Monkey Score is such a guardrail you apply yourself.
The studies at a glance
More trust in AI, less critical thinking
Survey of 319 knowledge workers. The more people trust generative AI, the less critical thinking they report; the more they trust their own skills, the more. Critical thinking shifts from doing to verifying, integrating and steering.
What it means for Monkey Score: Monkey Score makes visible whether answers are checked or simply accepted.
Frequent AI use goes along with less critical thinking, especially among younger people
Survey and interviews with 666 participants. Frequent use of AI tools correlates with lower critical thinking, mediated by “cognitive offloading”. Younger participants were more dependent. Note: these are correlations, not proof of cause and effect.
What it means for Monkey Score: That is why Monkey Score focuses on young people and rewards thinking for yourself.
AI without guardrails can harm learning
Field experiment with nearly 1,000 high school maths students. With a plain AI chat they did better on practice problems but worse afterwards without AI than peers who never had access. An AI tutor giving hints instead of answers largely prevented this effect.
What it means for Monkey Score: How you use AI matters. That is exactly what Monkey Score measures.
Less brain connectivity and recall when writing with ChatGPT
EEG study with 54 participants writing essays with ChatGPT, a search engine or no tools. The ChatGPT group showed the weakest brain connectivity and remembered their own texts less well. The authors call this “cognitive debt”. Caveat: small sample, not yet peer-reviewed.
What it means for Monkey Score: A hint that active thinking matters when working with AI.
Students often let AI deliver answers directly
Analysis of anonymised student conversations with Claude. About half were “direct” (asking for an answer or finished text), the other half “collaborative” (working things out together). Demanding thinking tasks such as creating and analysing were often handed to the AI.
What it means for Monkey Score: This distinction mirrors the response types in Monkey Score: accepting or thinking along.
Source: Anthropic (2025): Anthropic Education Report: How university students use Claude.
Small thinking prompts reduce blind trust in AI
Experiment with 199 participants: deliberate prompts (“think first, then ask the AI”) reduced overreliance on AI, but people liked them less.
What it means for Monkey Score: Monkey Score therefore uses voluntary, playful incentives instead of force.
How Monkey Score measures
Monkey Score is not a scientifically validated test but a transparent mirror based on fixed rules. Each interaction is measured: one AI answer and your response to it.
1. Reading
Each answer is split into sections. Monkey Score measures how long each section was visible on screen (viewport time, an established method in reading research) and derives a reading speed in words per minute. Adults read about 240 words per minute on average.
- up to 350 words/minute: read
- 351 to 800: skimmed
- faster, or visible for less than 1 second: scrolled past
2. Responding
Your next message is classified locally into a response type: waving through, asking for instructions, wanting to understand, questioning, or own contribution. The text is analysed only at the moment of sending and discarded immediately.
3. Points
Reading and responding together give 0 to 100 points. Example: read carefully and questioned = 100, scrolled past and waved through = 0. The score covers the last 7 days, with recent interactions weighted more.
Limitations
The measurement is based on behaviour, not on understanding itself. Scrolling slowly is not the same as reading. Monkey Score is therefore meant as a prompt for self-reflection, not as a grade or a way to judge other people.
Sources for the method: Lagun, D. & Lalmas, M. (2016): Understanding User Attention and Engagement in Online News Reading. WSDM ’16. · Brysbaert, M. (2019): How many words do we read per minute? A review and meta-analysis of reading rate. Journal of Memory and Language 109.