Hacker News Reader

HNHacker News

Top New Best Ask Show Jobs

Comment by xr4ti | Hacker News Reader

Full thread

xr4ti·

This looks like something that is just begging to misinterpreted and abused by investigators and prosecutors.

>Opaque technology with an unknown error profile? Check.

>Broad search criteria that guarantee a high false positive rate? Check.

>Evidence that requires careful consideration of prior odds to properly assess? Check.

Imo, the main ingredients here were the low population density and the suspect's own sense of guilt/fear.

View on HN

I'd be curious to see a comparison between the accuracy of this new technology and good old-fashioned eyewitnesses (who are working from memory) and hand sketches done by a human artist. I wouldn't assume that this is less accurate than the human approach.

Eyewitnesses are potentially quite unreliable and the reliability can be heavily influenced by the stress level of the situation, relative racial experience of the witness (It's harder to recognize/differentiate members of an ethnicity you don't spend much time around), interview techniques and line-up methodology.

Yet eyewitness testimony is generally treated as highly accurate and some places do not allow expert witnesses who can educate the jury on the factors that can influence the reliability of eye witness testimony.

Eyewitnesses can easily be confronted and cross examined. Questioning eyewitnesses is a relatively common skill among defense attorneys, and the arguments being made in such cross examinations are easily grasped by jurors.

The algorithm behind something like this might not even be made available to defense counsel. If it is, examination will require an expert witness (which costs $$$$), and it's doubtful that 12 average citizens would be able to assess the validity of or even comprehend the arguments.

Reply on news.ycombinator.com