← All interviews

Interview · Forensic science

Hal S. Stern

Provost and Executive Vice Chancellor, Distinguished Professor of Statistics, UC Irvine; Co-director, CSAFE

https://youtu.be/_gL8jY8MghA ↗

Chapters

Key moments from the conversation, each paired with a Korean case that raises the same question.

Chapter 01

Open at this moment ↗

Does matching forensic evidence actually prove guilt?

  • Early research looked at bullet-lead analysis, which compared the chemical make-up of bullets found at crime scenes with bullets belonging to suspects.
  • Two bullets can be chemically near-identical while hundreds of thousands of other bullets were manufactured from the same batch.
  • So a match does not by itself establish a meaningful link between the suspect and the crime.

The Korean parallel

Hwaseong eighth murder case (1988)

Yoon Seong-yeo was convicted partly on forensic hair analysis said to connect him to the scene. Problems with that analysis later emerged, and he was acquitted at a 2020 retrial after spending 20 years in prison.

Chapter 02

Open at this moment ↗

Is forensic science always scientifically proven?

  • DNA analysis rests on a substantial scientific foundation that was developed independently of criminal investigations.
  • Other methods, fingerprint comparison among them, were used in court for decades before rigorous studies tested their reliability under realistic conditions.
  • The 2009 National Academy of Sciences report and the 2016 PCAST report both challenged the scientific foundations of several forensic disciplines.
  • Scientific validity has to be earned through controlled testing, not assumed from long use.

The Korean parallel

Korean bloodstain-analysis research (2025)

A study of 188 Korean criminal judgments from 2012 to 2023 found cases in which courts rejected bloodstain-pattern conclusions because of mistaken assumptions or inadequate measurements.

Chapter 03

Open at this moment ↗

Why fingerprints are less certain than they look

  • Fingerprint identification is often seen as an objective process that mechanically settles who someone is.
  • Crime-scene prints are rarely as clear as prints taken directly from a suspect, so examiners routinely compare a clean print against a distorted one.
  • How reliable that comparison is depends on how much usable information the print holds, which varies from case to case and leaves room for the examiner's judgement.

The Korean parallel

Busan partial-fingerprint case

A man received a five-year sentence after a partial fingerprint on a knife was linked to him. Four experts supported the identification, but the defence questioned whether the knife was used in the crime and whether the print identified the perpetrator. The conviction was being challenged; reporting did not establish that the identification itself was mistaken.

Next interview

Michael Parente →