- AI algorithms already shape bail, sentencing, and parole decisions in courts.
- Human judges are inconsistent — sports results can influence sentence lengths.
- AI lacks embodied experience, making it unsuitable for full leadership roles.
Professor Hannah Fry's favorite example of judicial inconsistency: Judges tend to give harsher sentences in towns where the local sports team has lost recently.
During her appearance on This Morning, she explains why artificial intelligence in courtrooms might actually improve justice - despite our instinctive recoil from the idea.
The mathematician and TV presenter, promoting her new show "Infinite Explorer" on National Geographic, tackles three provocative AI headlines with characteristic insight.
Key figure
2035
The year Fry predicts we may be closer to AI companions than to human friends
The Inconsistency Problem
Apparently AI is already being used in courtrooms in the US and UK. Algorithms propose bail, sentences, and parole assessments, but the de facto decision still rests with the judge.
What is a risk-assessment algorithm?
A risk-assessment algorithm is a software tool that scores a defendant's likelihood of reoffending or failing to appear in court, based on factors like criminal history and background data. Judges can use the score as one input among many when deciding on bail, sentencing, or parole – but the legal decision remains theirs. Tools like COMPAS are among the most widely used in US courts.
Fry's initial reaction mirrors most people's: "You don't really want a kind of cold faceless machine involved in decisions about human freedom."
But human judges aren't consistent either.
Take the same case to different judges, you get different decisions. Take it to the same judge on different days, same problem. The randomness is enormous.
AI Leaders and Emotional Intelligence
When asked about AI prime ministers–a real proposal from a Japanese political party–Fry highlights a crucial limitation. Current AI systems excel at analysis and distilling information, but they're fragile in other ways.
"It might have read all of the internet, but it doesn't understand what it feels like to stand outside on a cold, rainy day," she explains. "It hasn't lived in the world in a human body."
Leaders need emotional connection and empathy. For now, human-AI teams work better than replacing humans entirely.
The Junk Food Relationship
The most concerning prediction: we'll be closer to AI than friends by 2035. Fry sees this already happening through chatbots that offer constant validation without reciprocal demands.
Her analogy cuts deep: "It's like the junk food version of human relationships. It might taste like food, but it doesn't mean that it's good for you."
Real relationships require showing up for others as much as they show up for you.
Fry's perspective balances pragmatic acceptance with human-centered caution. AI can improve flawed human systems, but replacing human judgment and connection entirely risks losing something essential about what makes us human.
Fact Check: Claim-by-Claim Verification Verified
All factual claims verified against the original This Morning interview, National Geographic show listings, and independent sources. Interview quotes, AI courtroom usage, and Japanese AI PM proposal all confirmed.
Commentary
- This article is primarily an interview summary reporting Fry's opinions and predictions, which are inherently subjective rather than factual claims.
- The 2035 prediction was a headline Fry responded to on This Morning, not her own precise forecast. The article could be clearer about this attribution.
- COMPAS and similar tools remain controversial due to documented racial bias issues, which the article doesn't address in detail.
Sources used for verification
Academic/Peer-reviewed:
- Machine Bias in Criminal Sentencing - ProPublica
- AI in UK Justice System - Oxford Tech & Justice
Other reliable sources:
- This Morning interview - YouTube
- Infinite Explorer - National Geographic
- AI Leader proposal - Japan Times
- AI chatbot analysis - Science Focus
Fact-checked by Perplexity Sonar Pro on 2026-03-14
