Graham Pike and Camilla Elphick from the Open University’s Centre for Policing Research and Learning have published new research on the phenomenon of the websleuth.
You’ve just witnessed a crime. It happened quickly, too quickly for you to film on your phone, but you got a good look at what happened. You gave the police a statement at the scene and said you would be happy to help further if needed. The next day you can’t stop thinking about what happened and find yourself in front of your laptop, looking for any news stories or further information on the event. It was just a theft at a party, so there are no headlines on the BBC, but you notice one of your friends has posted something on social media. Looking further you find lots of pictures from the party as well as further comments, check-ins and people tagged… and find yourself wondering if someone happened to capture an image of the perpetrator, or maybe someone even knew who they were and tagged their face! You look at image after image and follow link after link. You have become a websleuth.
A few years ago, our research team met with detectives and staff from Thames Valley Police to share ideas about how best to obtain evidence from eyewitnesses. The police said that one of the biggest issues facing them was what to do with a witness who showed up at the station with a picture of the perpetrator they had found on social media. According to their codes of practice, that witness still needed to be shown a formal (video) identification parade to see if they could identify the perpetrator. Usually a witness would compare the faces in the parade to their memory of the perpetrator from the crime, but the question was whether a witness who had found an image on the web would instead use this as their target face. To put it another way, if a witness found an image of the wrong person on social media would this over-write their memory of the real perpetrator, causing them to identify an innocent person from a formal identification parade?
Although there has been a great deal of research on the psychology of eyewitness identification (you may well have read about some in your studies), the specific question of whether formal identification evidence would be affected by a witness finding potential images of the perpetrator on social media had yet to be addressed. On one hand, research has found that eyewitness memory can be incredibly inaccurate and that introducing witnesses to information after they have seen a crime, known as ‘post-event’ information, can alter their memory. On the other, our own research (and that of others) has found that constructing a facial likeness, or ‘E-FIT’, does not seem to affect the witness’ memory of the perpetrator’s face, nor the accuracy of their subsequent identification decision at a parade (Pike et al., 2019a). This holds true even for contemporary systems that can produce lifelike images (Pike et al., 2019b).
Searching for, and particularly finding, an image on social media is obviously a different situation than working with the police to make a facial likeness. For one thing, the photo on social media will look exactly like the person, whilst E-FIT images tend to bear only a general resemblance. So, we set out to research what impact being a websleuth might have on the accuracy of eyewitness identification evidence. In designing the study, we instantly ran into a problem faced by the police, which is that social media has become quite popular and as a result there are rather a lot of images out there so how can you hope to ever know what a witness has actually looked at! It became clear that we could not use existing social media sites and so our only option was to build our own – that way we would know exactly what our participant-witnesses looked at and could control whether they saw an image of the perpetrator or instead an image of an innocent person that happened to look like the perpetrator. We called our site ‘FriendFace’.
In our experiment, participants were all shown the same video of a staged crime (a man stealing a bag at a party) and were then randomly allocated to one of three conditions. In one condition, the control condition, they did not look at FriendFace at all. In the other two conditions they were free to search FriendFace, although in one condition it contained images of the perpetrator whilst in the other these images were replaced with those of an innocent lookalike. Half of the participants in each condition were then shown a lineup containing the perpetrator (known as a ‘target present’ lineup) and half were shown a lineup that did not contain the perpetrator but did contain the innocent lookalike (known as a ‘target absent’ lineup).
Our results showed that participants in the lookalike condition were less likely to identify the perpetrator in the target present lineups than participants in the other two conditions, and that there were no differences for the target absent lineups. This suggests that being a websleuth who does not find the culprit may make a witness less likely to identify the perpetrator at a formal identification parade but may not make them more likely to identify an innocent person who they did see on social media. However, in our research the participants searched FriendFace immediately after seeing the crime and then waited a week before attempting to make an identification from the lineup, which raises the question of what might happen if the witness looked (or relooked) at an image of an innocent person on social media just before providing identification evidence. A question we plan to address in our next experiment.
To read more about this research, you can find the full-text of our article here:
Read about Graham Pike: http://www.open.ac.uk/people/gep34
Read about Camilla Elphick: http://www.open.ac.uk/research/people/ceje3
Read about the Centre for Policing Research and Learning: https://www.open.ac.uk/centres/policing/
Find out more about Citizen Forensics: