The triumphant march of facial recognition

Original article was published on Artificial Intelligence on Medium


The triumphant march of facial recognition

Last week, the Telangana State Council of Higher Education (TSCHE) announced that it will introduce a facial recognition system. With it, students should be able access their education records using a selfie and a few details. It seems like the days of treks to higher education departments and dharnas to get marksheets may be numbered.

TSCHE joins a movement of a number of organisations making facial recognition an everyday reality for Indians: airports facilitating boarding, governments identifying suspects during riots, companies identifying fraudulent customers, Gujarat monitoring teachers’ attendance, and Telangana piloting its use verify voters during elections.

As this movement grows, and the world becomes increasingly digital, the 20th century sci-fi vision of your face as your key will seem more real. All kinds of organisations building databases of faces. Already, if you’re online, chances are that your face is part of someone’s database. While facial recognition is convenient, quick, and helps you protect your identity, it also carries large risks, most of which are behind the lens.

Firstly, what if the algorithm gets it wrong? In the west, we’ve already seen cases of people being wrongly arrested because of an incorrect facial match. These errors are more likely for vulnerable groups that are under-represented in the training algorithms. In the US, that is often African-Americans; in India, it could be lower castes or scheduled tribes. And what happens if these groups do get affected? In a country where consumer and citizens struggle to enforce their rights anyway, will they be able to seek quick redress against such errors?

Secondly, even if the algorithm is correct, what happens when your face is analysed without your knowledge or consent? For example, researchers have used it to predict sexual orientation — what does this mean for sexual minorities among us who may be closeted? Or when facial recognition is used to constantly track and monitor you, including at public events?

The goal for us is to retain the benefits while minimising the harms. As researcher Smriti Parsheera argues in her paper, facial recognition has proliferated in India without adequate safeguards and regulation. The data privacy bill is an important step ahead, but insufficient. We need to have an ongoing conversation about when, where and why we use facial recognition. What recourse do we have if they feel aggrieved? In a country that has a tradition of controlling access to faces, through both the ghunghat and niqab, we deserve to know what happens with digital versions of our face.

So, next time you’re asked for a selfie as a key, ask how your data will be used.

P.s. Learn more about facial recognition through this short John Oliver episode.