[Barbing]

>things you see with your own eyes

Oh I wouldn't trust myself!

  In his study, "Nearly 2,000 witnesses can be wrong", Buckhout performed an experiment with 2,145 at-home viewers of a popular news broadcast. The television network played a 13-second clip of a mock robbery, produced by Buckhout. ... The people at home could call a number on their screen to report which suspect they believed was the perpetrator. The perpetrator was suspect number 2. ... Approximately equal contingents of participants chose suspects 1, 2, or 5, while the largest group of participants, about 25 percent, said they believed the perpetrator was not in the lineup. Even police precincts called in and reported the wrong man as the one they believed committed the crime.

https://en.wikipedia.org/wiki/Eyewitness_testimony#Reliabili...

Also-mentioned concepts: misinformation effect, source misattribution, confirmation bias.