[noduerme]

Wouldn't proximity and time be rather important? Seeing someone briefly from a distance, in a traumatic and stressful moment, could easily lead to mistaken identification. On the other hand, the Nuremberg trials relied on eyewitnesses to identify people guilty of genocide who were trying hard to disguise their appearance, and this was possible because the witnesses had been in relatively close proximity to the accused for a long period of time.

Just like an AI can't make a perfectly accurate 3D replica of a face from a single 2D image, the human brain can't. But that doesn't mean it can't get closer and closer with more data.

It seems to me that the real problem is not eyewitness testimony itself, but the methods used to validate the degree of certainty behind it.

[nvrspyx]

I think the parent agrees, hence:

> when the person in question was not previously known to the witness

[noduerme]

What I'm taking issue with is the word "previously".

"Previously" implies a family member or neighbor. But if you were caught and tortured by someone you didn't previously know for awhile, you would still probably identify them pretty accurately.

[boomboomsubban]

>But if you were caught and tortured by someone you didn't previously know for awhile, you would still probably identify them pretty accurately.

If that was your only interaction with them, I doubt it. You're likely in the most stressful situation of your life, memorizing what a person looks like may not be high on your list of priorities.

There's the famous Ronald Cotton case where a woman was raped and consciously tried to memorize their attackers face. She proceeded to pick an innocent man out of two lineups until a decade later DNA evidence proved his innocence. The actual perpetrator was also in the lineup, and she swore she had never seen him.

[londons_explore]

There is always the possibility the real perpetrator said "don't help the police catch me. If you do, then my friends will murder your family.".

Many of the cases of mistaken memory might simply be 'deliberate forgetting'

[boomboomsubban]

If you're threatened the "right" answer would be to say you don't recognize anyone in the lineup. Not accuse an innocent person.

And we know of many cases where eyewitness accounts are wrong purely because human memory kinda sucks. Maybe some people are lying about what they know, but that doesn't make eyewitness accounts more reliable.

[noduerme]

The logical endpoint of this argument is that no one can be convicted for genocide without DNA evidence, regardless of how many eyewitnesses saw that person executing people and can identify them on sight.

I find this unacceptable.

[edit] There's something called a preponderance of evidence. One eyewitness may or may not be able to provide that alone, e.g., "he has a tattoo of a dolphin on his penis". Multiple witnesses in tandem can add to the preponderance of evidence based solely on eyewitness accounts. Yes, it's worth it to free ten criminals to prevent convicting an innocent person; but if we accept that as a starting point, it's also immoral to discount witnesses or victims. Using one person who wrongly accused someone of rape to corroborate the idea that people are incapable of identifying the people who raped them is, frankly, gaslighting rape victims.

[cyphar]

> The logical endpoint of this argument is that no one can be convicted for genocide without DNA evidence, regardless of how many eyewitnesses saw that person executing people and can identify them on sight.

There are other forms of evidence that a genocide took place other than eyewitness testimony (orders, technical plans, the testimony of subordinates, physical evidence of mass killings). You would hope that a trial on crimes against humanity would have more than just eyewitness testimony.

> I find this unacceptable.

I also find it unacceptable for someone to be convinced based entirely on circumstantial evidence and eyewitness testimony, only to discover they were innocent decades later.

> [edit] There's something called a preponderance of evidence.

"Preponderance of the evidence" (meaning that it is more likely than not) is a lower bar than "beyond reasonable doubt" (which means what it says on the tin). What you described (having multiple corroborating witnesses) is actually not what preponderance of evidence means.

[nvrspyx]

I think this is unnecessarily pedantic. If the crime was over a prolonged period of time or on multiple instances with the same witness present, then you eventually become familiar with the perpetrator.

The original commenter obviously meant that a witness should be reasonably familiar with the person, whether that be from knowing the person prior to any crime or, in rare cases like kidnappings/torture/etc., during the prolonged period of time with the accused. It's not like we're writing legislature ourselves here; those rare cases probably just didn't come to mind when the original commenter wrote their comment. Informal internet commentary need not be so needlessly scrutinized.

But if we really want to be pedantic with your example, you would become familiar with the person from the "previous" torture sessions.