Ian Forrester over at Cubic Garden has submitted a GDPR request to ClearView AI, the alt-right linked company that is hawking their facial recognition database (based on scraped online images) to law enforcement as well as commercial outfits. Should be interesting to follow along. Recently IBM stopped facial recognition work (they previously showed not being up to speed with CC licensing it seemed to me), and others like Amazon and MicroSoft did too when it comes to law enforcement. Facial recognition is extremely sensitive to bias.

facial-recognition-1Facial recognition 1, by EFF, license CC BY

Some disturbing key data points, reported by the Guardian, from a Congressional hearing in the US last week on the usage of facial recognition by the FBI: “Approximately half of adult Americans’ photographs are stored in facial recognition databases that can be accessed by the FBI, without their knowledge or consent, in the hunt for suspected criminals. About 80% of photos in the FBI’s network are non-criminal entries, including pictures from driver’s licenses and passports. The algorithms used to identify matches are inaccurate about 15% of the time, and are more likely to misidentify black people than white people.” It makes you wonder how many false positives have ended up in jail because of this.

At GEGF2014
Me, if you look closely, reflected in an anonymous ‘portrait’ (part of an exhibit on Stalinist repression and disappearances in Kazakhstan, 2014)

I am in favor of mandatory radical transparency of government agencies. Not just in terms of releasing data to the public, but also / more importantly specifying exactly what it is they collect, for what purpose, and what amount of data they have in each collection. Such openness I think is key in reducing the ‘data hunger’ of agencies (the habit of just collecting stuff because it’s possible, and ‘well, you never know’), and forces them to more clearly think about information design and the purpose of the collection. If it is clear up-front that either the data itself, or the fact that you collect such data and in which form you hold them, will be public at a predictable point in time, this will likely lead to self-restraint / self-censorship by government agencies. The example above is a case in point: The FBI did not publish a privacy impact assessment, as legally required, and tried to argue it would not need to heed certain provisions of the US Privacy Act.

If you don’t do such up-front mandatory radical transparency you get scope creep and disturbing collections like above. It is also self-defeating as this type of all encompassing data collection is not increasing the amount of needles found, but merely enlarging the haystack.

Using tech to flip facial recognition in video stories from Iran, at SXSWi
image by Sheila Scarborough, CC-BY