AI summary by TwinMind
The Math Problems That Trick Smart People
Veritasium
14m 35s
Research Paper Overview
- Paper shows particular problem type where smarter people more likely to get wrong answers
- American friend Wy conducted street interviews replicating study findings
- Lead author: Dan Kahan, studies science communication
- Sample: 1,111 nationally diverse Americans
Skin Cream Study Design
- Fictitious study: people with skin rashes divided into groups
- Experimental group: used new cream for 2 weeks
- Control group: used no cream for same period
- Measured how many rashes improved vs worsened in each group
- Results presented in table format
Initial Skin Cream Results
- Intuitive answer: cream worked (biggest number showed improvement in cream group)
- Correct proportional reasoning analysis:
- Cream group: ~3x more people got better than worse
- Control group: ~5x more people got better than worse
- Conclusion: cream actually made rashes worse on average
Numeracy Assessment & Results
- Participants tested on numeracy before answering questions
- Numeracy: capacity to reason well about quantitative information (not complex math)
- Numeracy scores roughly normally distributed across sample
- Higher numeracy = higher accuracy on skin cream question (expected result)
- Version 1: cream made rashes worse (as described above)
- Version 2: same table with reversed column headings (cream actually helped)
- High numeracy groups most accurate
- Below numeracy score 5: results flat (people expect cream to help)
Political Dimension Introduction
- True purpose: test if politics affects reasoning ability
- Skin cream data separated by political affiliation (liberal Democrats vs conservative Republicans)
- No difference between groups - same pattern regardless of party
- Makes sense: skin cream completely apolitical topic
Gun Control Study Variation
- Other half of participants saw identical data structure
- Topic changed to gun control study instead of skin cream
- Background context:
- Democrats generally believe gun control reduces crime
- Republicans believe gun control increases crime (takes guns from good guys)
- Study design: cities with/without concealed carry bans, crime rates monitored
- Two variations: one where gun control decreased crime, one where it increased
Political Bias Results
- Republicans shown "gun control increases crime" table: accuracy improved with numeracy (normal pattern)
- Republicans shown "gun control decreases crime" table: numeracy no longer affected accuracy
- Numeracy score 7-8 barely more accurate than score 0-1
- Democrats showed reverse pattern:
- High accuracy when data showed gun control reduced crime
- Numeracy didn't help when data showed gun control increased crime
- Same exact data/question as skin cream - only difference was political context
Magnitude of Political Bias
- Low numeracy people: 25 percentage points less likely to get right answer when threatening to ideology
- High numeracy people: 45 percentage points less likely to get right answer when contradicts beliefs
- Rather than using math ability for correct conclusion, selectively applied to justify pre-existing beliefs
Broader Pattern Confirmation
- Same effect observed with other measures:
- Science literacy: higher literacy = more polarization
- Active open-minded thinking: more proficiency = more polarization not less
- Pattern extends to fracking, global warming issues
- As numeracy increases, people more divided about gun control efficacy/risks
Street Interview Insights
- People claim to be rational, open to changing beliefs based on evidence
- Reality: continue holding partisan beliefs regardless of evidence while believing they reached conclusions independently
- Humans believe what fits with their tribe - highly rational for social survival
- Throughout evolution, depended on groups for survival
- Being ostracized almost as bad as walking off cliff
- Questioning status quo jeopardizes place in society
Potential Solutions from Kahan
- Avoid partisan rhetoric and loaded terms
- Focus on specific local policies without buzzwords
- No villainizing other side, just constructive data-driven solutions
- Example: Southeast Florida bipartisan lawmakers on sea level rise
- Don't debate if climate change man-made
- Just address existing resident challenges
- Foster curious mindset over ideological thinking
- Science curious people more willing to examine all evidence including contradictory
- With increasing science comprehension: usually more polarization
- With increasing science curiosity: polarization pattern not observed
Ground News Sponsor Segment
- Organization tackling polarization and information echo chambers
- Gathers news sources worldwide, shows political leanings
- Example: FBI violent crime statistics
- Left-leaning outlets: focused on 3% drop in violent crime
- Right-leaning sources: suggested crime increased under Harris/Biden
- Different data sources: FBI statistics vs DOJ survey
- FBI included fatal crimes, excluded unreported crimes
- DOJ survey excluded fatal crimes, included unreported crimes
- Personal bias dashboard tracks individual blind spots
- 50% discount on Vantage plan through link
