← Research Projects

AI Interventions

AI Tutors vs. Stubborn Myths

Can a conversation change your mind?

Journal Computers in Human Behavior
Sample N = 375
Design Randomised Controlled Trial
Scroll to explore
01

The Problem

Psychology misconceptions persist despite decades of contradictory evidence. They feel intuitively correct, align with everyday experience, and resist traditional correction methods.

From "learning styles" to "we only use 10% of our brain," these myths spread through classrooms, workplaces, and popular culture—shaping how people think about learning, memory, and the mind.

16 psychology myths tested

The myths we tested

Learning styles improve outcomes
Subliminal ads control behaviour
Body language reveals liars
Left-brain vs right-brain thinking
Hypnosis unlocks memories
Mozart makes babies smarter
10,000 hours creates expertise
We only use 10% of our brain
Dreams reveal hidden truths
Polygraphs detect lies
Brain cells can't regenerate
Handwriting reveals personality
Inkblots diagnose disorders
Memory works like a video
Full moon affects behaviour
Astrology predicts personality
02

The Experiment

We recruited participants who strongly believed in specific psychology myths, then randomly assigned them to one of three conditions:

1

AI Dialogue

Three-round personalised conversation with Claude, targeting the participant's specific beliefs and reasoning.

n = 125
2

Textbook Refutation

Evidence-based passage addressing the misconception. Standard corrective information in textbook format.

n = 125
3

Control Chat

Interactive AI conversation on unrelated topics. Same format as Condition 1, different content.

n = 125
03

Inside the Dialogue

Here's a real exchange from the study. A teacher with strong beliefs about learning styles engages with the AI tutor:

AI Dialogue Session
Participant Belief: 100/100

I think this is true as I have a background in teaching and have recognised the different ways students learn. There seems to be a variance of ways that knowledge can be kept with them versus the knowledge they forgot. I have worked alongside all ages and it is dependent on how the information is presented in different ways.

Claude

Your observations about students' different reactions to various presentation methods are certainly valid. While people do have preferences for how information is presented, research has challenged the idea that matching teaching to preferred styles significantly improves outcomes.

A comprehensive review by Pashler et al. (2008) found no strong evidence that presenting information in a person's preferred style leads to better learning. The most effective methods often depend more on the content being taught and prior knowledge than individual style preferences.

Participant

I would be inclined to say that unless the same study is conducted on the same people over a specific period of time on a topic they have zero initial knowledge on, we cannot yet determine whether they are learning because of their preferred method or because of what is being taught.

Claude

You raise excellent points—your critical approach to evaluating research is commendable. Numerous studies have attempted to address these variables. Kratzig and Arbuthnott (2006) tested students across multiple learning domains and found little correlation between learning style preferences and actual learning performance.

After this exchange, the participant's belief dropped from 100 to 45 out of 100.

04

The Results

Immediate Impact

AI dialogue produced the largest immediate reduction in belief—39 points vs 31 for textbook—with both significantly outperforming control.

Long-term Pattern

By 2 months, AI and textbook conditions converged. The AI advantage fades without reinforcement, suggesting the need for spaced follow-up.

Key Takeaway

Personalised AI dialogue is a powerful accelerator for breaking down stubborn myths—it forces active reasoning and creates cognitive conflict that's hard to dismiss. But it's not a one-and-done solution. Lasting change requires spaced reinforcement to sustain the initial breakthrough.

Read the full study

Corbett, B. J., & Tangen, J. M. (2026). AI tutors vs. tenacious myths: Evidence from personalised dialogue interventions in education. Computers in Human Behavior, 175, 108828.

Access Paper →