#87 Hannah Fry: The Role of Algorithms

Episode Summary

Executive Summary: In this episode of The Knowledge Project, host Shane Parrish interviews mathematician and author Hannah Fry about making math relevant, the human element in algorithms, and lessons from applying math to relationships. Fry shares how early exposure and practice sparked her love for math, discusses why math education must show real-world applications, and critiques the blind trust placed in algorithms. She uses vivid examples—from autonomous cars to recidivism risk assessment—to argue that algorithms must be designed with human psychology in mind, not as replacements for human judgment.

Main Topics: Making Math Engaging for Students (Priority: 5/5): Fry argues that math education fails to show how math underpins modern life, making it invisible and uninteresting. She advocates for revealing the human stories behind mathematical discoveries and demonstrating practical applications to combat the 'I'll never use this' attitude. Human-Machine Interaction and Trust in Algorithms (Priority: 5/5): The conversation explores how algorithms influence decision-making, from GPS navigation to criminal sentencing. Fry stresses that algorithms must be designed to allow human oversight, not to replace it, using examples like Japanese tourists driving into the ocean and judges blindly following recidivism risk scores. Medical AI and Overdiagnosis (Priority: 4/5): Fry highlights how overly sensitive cancer detection algorithms can lead to unnecessary, invasive treatments. She explains that many tumors are harmless, and perfect detection can cause more harm than good, emphasizing the need for transparent, interrogable AI systems in medicine. Algorithms in the Judicial System (Priority: 5/5): Using the case of Christopher Drew Brooks, Fry shows how recidivism prediction algorithms can be illogical and biased. She argues that while human judges are inconsistent, blindly following algorithms is dangerous, and advocates for a balanced, transparent approach to algorithmic support in courtrooms. Mathematical Models During Pandemics (Priority: 4/5): Fry discusses her collaboration with the BBC to improve data on human movement, which now informs COVID-19 models. She explains how exponential growth is counterintuitive and why math is essential for policy decisions when pharmaceutical interventions are unavailable. Applying Math to Romantic Relationships (Priority: 3/5): Fry humorously applies optimal stopping theory to dating, advising people to explore 37% of options before committing. She also discusses John Gottman's work on marital arguments, revealing that couples with low negativity thresholds—who address issues immediately—have more successful long-term relationships. The Story of Galois and Humanizing Math (Priority: 3/5): Fry recounts the tragic story of mathematician Évariste Galois, who spent his last night before a duel frantically writing down his mathematical theories. This story illustrates how attaching human narratives to math can make it more engaging and memorable.

Key Arguments: Math education must demonstrate real-world relevance to engage students, not just teach abstract rules. Algorithms should augment human decision-making, not replace it; transparency and human oversight are crucial. Exponential growth is deeply counterintuitive; better understanding can improve public response to pandemics. Medical AI must be designed to allow interrogation by doctors to avoid overdiagnosis and unnecessary treatments. In judicial systems, algorithms can reduce inconsistency but must be carefully monitored to avoid amplifying bias. Human stories and narratives are essential to make math appealing and accessible to broader audiences. Applying mathematical models like optimal stopping theory to dating can offer useful, albeit imperfect, heuristics.

Data Points: Optimal stopping percentage for dating: 37% - Fry explains that mathematically, you should explore the first 37% of potential partners before committing to the next best one. Rice on chessboard total: 18 quintillion grains - Fry uses the classic example to illustrate exponential growth: enough rice to cover the area of Liverpool stacked 3 kilometers high. COVID-19 doubling time in early 2020: Every five days - Fry references the virus's doubling rate to explain exponential growth and the need for early intervention. Number of people in UK travel survey before BBC project: 1,000 - Fry notes that prior to her project, the best UK data on human movement came from a 2006 survey of only 1,000 people. Deaths in UK at time of recording: ~150 - Fry mentions that despite low current numbers, mathematical models predicted a much worse situation, driving policy. Age difference in recidivism algorithm example: 19 vs 36 years old - Fry gives an example where a 19-year-old was considered high risk due to age, while a 36-year-old committing the same crime would be judged low risk.

Pivotal Quotes: "I think that the maths itself needs to be humanized if it's to properly fit in with our society." — Hannah Fry: Fry explains why algorithms must account for human psychology and context, not just raw data. "To fly a plane, you need three things: a computer, a human, and a dog. The computer is there to fly the plane, the human is there to feed the dog, and the dog is there to bite the human if ever it touches the computer." — Hannah Fry: Fry uses this joke to illustrate that humans should not interfere with algorithms in high-stakes, fast-moving systems. "I think one of the big complaints that you get from school kids is like, 'Well, I'm never going to use this stuff, what's the point of it?' And I think really showing just how dramatically important maths is to virtually every aspect of our modern world, I think that that's something that can really make the subject come alive." — Shane Parrish (quoting common student complaint) and Hannah Fry (endorsing the solution): This exchange frames the central challenge of math education and Fry's proposed remedy.

Implications: Listeners should recognize that algorithms are tools, not oracles. Effective use requires critical thinking, transparency, and human-centered design. In education, embedding real-world applications and human stories can transform math from a chore into a powerful lens for understanding the world.