The synopsis:
"Where people look when watching video evidence varies wildly and has profound consequences for bias in legal punishment decisions"
However, their study is a very narrow, statistically uncommon, use of video surveillance.
The NYU / Yale study found:
"Participants who viewed videotaped altercations formed biased punishment decisions about a defendant the more they looked at him. Participants punished a defendant more severely if they did not identify with his social group and punished him less severely if they felt connected to the group—but only when they looked at the defendant often."
This sounds reasonable / plausible, especially since the "video depicting an actual altercation between a police officer and a civilian, in which officer wrongdoing was ambiguous."
So, in circumstances, where the video does not draw a clear picture itself, human subjectivity in interpreting the actions in a video will become clear.
Of course, most video surveillance used as evidence is much more clear cut, showing that someone was at a premise at a certain time, etc.