An update on election Hackday and the goals of WhoTargets.me

Earlier in August, I went to a hack day with a few friends to work on WhoTargetsMe, a project started by some people in London. We ended up working on the project because we felt that platforms like Facebook had emerged, that were were powerful in the same way that you might consider TV and the press to be powerful when it came to influencing elections.

Thing is, there’s not much in the way of oversight for platforms like Facebook, especially during elections, so it’s very hard to see if the platform is being used in a malicious way.

The appeal for WhoTargetsMe, for me at least was that it was a clever approach to build a dataset to allow for some kind of scrutiny over how Facebook was being used in elections, and it seemed a good way to work towards the things Tom Steinberg outlined:

What I want is this: I want Facebook and Google to show goodwill by voluntarily publishing data on the political adverts that are purchased on their platform and shown to users in the U.K. in the next six weeks.

To be more specific, I want:

  • A copy of each unique advert (e.g image/text/video)
  • Data on who this advert was targeted at (e.g everyone/only women/only people in London)
  • Data on how many people have been shown each advert
  • Information about who the buyer was

Earlier this week, in a briefing with Techcrunch, Facebook announced something that felt like progress towards this goal:

Facebook briefed TechCrunch on the changes that include hiring 1,000 more people to its global ads review team over the next year, and making it so anyone can see any ad run by any organization on Facebook instead of only the ads targeted to them.

So, one of the key ideas about making adverts less ‘dark’ looks to like it might actually be delivered.

Also, largely as a result of more and more Russian interference in the election, Facebook agreed to share a set of ads with US Congress, and some information about it’s use. This snippet from their own blog is enlightening, but the underlying data doesn’t seem to have been shared beyond the congress investigation :

Most of the ads appear to focus on divisive social and political messages across the ideological spectrum, touching on topics from LGBT matters to race issues to immigration to gun rights. A number of them appear to encourage people to follow Pages on these issues.

Here are a few other facts about the ads:

  • An estimated 10 million people in the US saw the ads. We were able to approximate the number of unique people (“reach”) who saw at least one of these ads, with our best modeling

  • 44% of total ad impressions (number of times ads were displayed) were before the US election on November 8, 2016; 56% were after the election.

  • Roughly 25% of the ads were never shown to anyone. That’s because advertising auctions are designed so that ads reach  people based on relevance, and certain ads may not reach anyone as a result.

  • For 50% of the ads, less than $3 was spent; for 99% of the ads, less than $1,000 was spent.

It’s a shame that it takes a train-wreck of an election for this to come out, and I hope it means that for this kind of information to be shared, it doesn’t take the same kind of disastrous election we saw in November 2016. But at least it sets a precedent making it easier to campaign for this information to be shared more regularly, as part of a step to make the use of platforms like Facebook in elections more transparent.