#sxswapac

By: Riku Vassinen

Facial recognition is a new data point which enables brands to serve customers better driven by the latest mobile operating systems. You can pay with a smile or easily classify your photos with different people. Increasingly more companies are asking permission to recognize your face.

Your face is more than your address or other data points. In many ways it is more than a photograph of you. With facial recognition you can actually track person´s movements. This much more personal data point brings opportunities but there are big risks to it as well.

Recent MIT study found out that face recognition algorithms work for white men but not as well with black woman. There are two kinds of bias going on in here: training data and implicit bias.

Training sets of faces are generally constructed from people like your colleagues or your university mates. Then there might also be implicit bias by makers of algorithm, who quite often happen to be white male. Facial recognition company Kairos CEO Brian Brackeen announced in SXSW that they are creating a diverse dataset with all the races and genders and they will have it opensource so everyone can use it. 

Facial recognition can be used (and is already used in China) to track you. With countries with many CCTV cameras , government could track (maybe already tracks) your movements and if you stray from your normal moving patterns too much, they could identify those fluctuations as potential terrorist threat (usually connected with racial profiling). Your face also reveals other things than your identity. Programs can detect your gender, age, race and even certain diseases you have. You do not necessarily want companies know this information while you are using face recognition to pay for your hamburger more quickly. There is now big focus in universities to create programs that enable you to use face recognition for your identity, but not to detect other aspects about you:

In the above example the face can be used to identify you but not necessarily detect your other traits.

No algorithm is 100% perfect and there will be errors with facial recognition. In Florida there is currently on-going appeal against failed facial identification. Anecdotally in Notting Hill Carnival there was 35 identifications but only 2 of those were accurate. As the development has been really rapid in this pace, there are no existing legislation or regulations around the use of face data. Companies are also actively fighting against regulations. Governments are the biggest users of facial recognition and some governments have been quite proactive in this place. 

Facial recognition and other biometrifics bring huge opportunities to personalize your brand experience with users.

When we are planning to use facial recognition, we should also focus on algorithm quality: recognition accuracy (trying to minimize bias) and privacy (Notice and inform consent, preferably opt-in). Highlighted already in the previous post, users are weighting convenience over security. From brand perspective it is not zero sum game.