Bad science and bad statistics in the courtroom convict innocent people
Faulty science, dubious statistics and unreliable forensics are argued to play a major role in wrongful convictions, especially in systems like the U.S. where prosecutors are rewarded for high conviction rates. Commenters highlight how techniques from bite-mark analysis to GPS anklets, photo “enhancement,” medical age testing and shaken-baby diagnoses can be overstated or misunderstood by judges and juries with little technical or statistical literacy. Many see deeper structural problems—perverse incentives for police and prosecutors, underfunded defense, biased expert witnesses, and the heavy weight given to eyewitness testimony—and call for reforms ranging from neutral court statisticians to stricter standards and accountability for forensic evidence and experts.
Incentives and Misconduct in the Justice System
- Many comments argue the core problem is not just “bad science” but perverse incentives: prosecutors rewarded for convictions, police for arrests, prisons for occupancy, and agencies for revenue (fines, forfeiture, military gear).
- Selective enforcement and historical roots in social control are highlighted as systemic, not accidental.
- Others note that in practice prosecutors focus on winnable cases and often assume defendants are “guilty of something,” leading to overcharging and plea pressure.
Expert Witnesses, Forensics, and Statistics
- Multiple examples of overconfident or biased experts: ballistics, bite marks, blood spatter, bullet matching, shaken baby syndrome, medical age tests, and “100% certain” claims where empirical error rates exist.
- Prosecution experts are portrayed as especially prone to overstating certainty and facing few consequences; defense experts face harsher scrutiny and career risk.
- Some commenters report encountering very competent forensic specialists, showing quality varies.
- Several suggest courts and lawyers are deeply innumerate; bad statistics and probability reasoning persist despite decades of criticism.
Digital, Video, and Location Evidence
- GPS and electronic monitoring are described as technically fragile but routinely treated as precise. Examples include repeated false violation alerts and house-arrest arrests driven by bad signals.
- Geolocation, phone logs, and networking evidence are said to be easily misinterpreted or fabricated, with non-technical judges and juries vulnerable.
- Video “enhancement” (e.g., heavily processed stills, AI upscaling, JPEG/MPEG artifacts) is seen as ripe for abuse if the pipeline is not clearly explained.
Wrongful Convictions and ‘Innocence’ Rates
- Estimates cited for factually innocent people in prison range around 1–3% (some studies higher), implying tens of thousands of people.
- A much larger group—hundreds of thousands—are legally innocent but detained pretrial. Whether to count them as part of the “innocent behind bars” problem is debated.
Eyewitness Testimony and Cognitive Limits
- Eyewitness misidentification is repeatedly flagged as a major source of wrongful conviction, even in “face-to-face” crimes.
- Juries often treat eyewitness confidence as strong evidence; commenters argue human cognition prioritizes closure over truth.
Structural and Educational Issues
- Courts often exclude technically knowledgeable jurors or instruct jurors to defer to experts, undermining meaningful scrutiny.
- The profession is said to self-select for people weak in math and science, exacerbating misuse of statistics and technical evidence.
- Expert testimony is inherently adversarial; there is no built-in neutral arbiter of scientific validity.
Reform Ideas and Open Questions
- Proposals include: separating forensic labs from law enforcement; on-call neutral statisticians; stricter vetting and sanctions for experts; better transparency and punishment for prosecutorial misconduct.
- Some see doctrines like the exclusionary rule as essential to protecting rights; others view them as letting criminals go free while leaving police misconduct largely unpunished.
- A recurring unresolved question: what error rate in criminal adjudication is acceptable, and how much certainty should be demanded for different kinds of crimes?