Just over 3 years ago I wrote a post about 2 algorithms that appeared to be doing something that – if you or I did it – would break the law: Google’s image search algorithm and Twitter’s keyword search algorithm. At that time, they were displaying pictures of one of the killers of Jamie Bulger – someone who is not legally allowed to be identified under UK law.
This week, there was a very high profile story in the media about Adam Johnson, a footballer who was found guilty of a child sex charge. It is illegal in the UK to identify the victim of a sexual offence. There has been much said about individuals naming the victim in this case, but less said about the capability for algorithms to also do so.
Here is a quick look at whether Twitter & Google have managed to improve over the last 3 years, or whether there is still a chance they may inadvertently break the law in this way.
Adam Johnson’s name was one of the top trends in Twitter for much of the day. A click on the trending term took users to this search results page:
As you can see, there are some quite nasty ‘related searches’ displayed for his name. A click on the 2nd related search leads to this result:
I have blurred 2 entries there: A Twitter username whose account has since been deleted, and what appears to be a woman’s name. I do not know whether either of these is the victim, but it’s worrying that Twitter aren’t on top of suppressing these on such high profile trends.
(update: On checking several hours after publishing this post, and after ‘Adam Johnson’ stopped trending, some Twitter results have now been cleaned.)
Google’s algorithm fares a little better at first. It is only when reaching the foot of search results that their ‘related searches’ appear. Here are the results at the foot of the first search results page:
As you can see, several names are mentioned there. The first 2 names have been mentioned many, many times in the press – Adam Johnson’s former partner, and daughter. ‘adam johnson 15 year old’ is also present, but the results aren’t quite as nasty as the first set of Twitter ‘related’ searches. But, on clicking ‘stacey flounders’, and scrolling to related searches there, the following appears:
Again, as you can see, I have blurred the results partially, where Google lists a person’s name which cannot be explained by other means (2 of the other, unblurred names there have appeared in other sad news stories, and are explainable). The above is simply when clicking the name of Adam Johnson’s partner. When clicking the ‘adam johnson 15 year old’ related search, the following appears:
Again, I have blurred a result there. As you can see, all of the related results are quite nasty here, but the blurred one in particular is very worrying. Ie: As with Twitter’s algorithm, Google is specifically naming someone who may/may not be the victim in the case. Additionally, Dan Bell noted a similar issue appears in Google Image Search. As Google themselves should not know the name of the victim, again, it is worrying that a name is allowed to appear here. (I have attempted to notify them.)
It is over 3 years since I last wrote about this topic, where both Twitter’s & Google’s algorithms appeared to be displaying results which could break the law.
I do not know in the above examples if the names they display are the victim in this case (frankly I hope not). Either way, it is worrying that both Google & Twitter’s search algorithms seem still to be capable of doing something that would likely be illegal for any person in the UK to do. It is concerning both from the broad point of view of algorithms breaking the law (and causing harm to individuals), and from the narrower point of view of this individual case.