What our AI thought of us, at Melbourne Knowledge Week 2019

Biometric Mirror 30 may 2019

A head shot from Biometric mirror

So, what do the artificial intelligence systems programmable by computer systems today, think about their users? What do you think your computer would have to say about you?

The second exhibit from the Human-Computer Interaction (IDL) at Melbourne Knowledge Week was placed beside the Augmented Studio project, at the centre of the Meat Market gallery. Biometric Mirror stood tall, a curtained space that captured the intrigue of festival attendees. What lay inside was a provocative showcase of the facial analytical capabilities of an Artificial Intelligence (AI) algorithm created today.

The Biometric Mirror project was guided by an immensely experienced team of IDL researchers, within the School of Computing and Information Systems at The University of Melbourne. This project was a partnership between Microsoft and the University of Melbourne, through the establishment of the Microsoft Research Centre for Social Natural Users Interfaces on campus.

The multi-disciplinary team of scholars, artists, scientists and other professionals behind Biometric Mirror were led by Dr Niels Wouters, a Post-Doctoral Candidate at The University of Melbourne and Head of Research at the Science Gallery in Melbourne.

The project offered many festival attendees their first experience with “Sci-Fi technologies” intermingling with everyday realities. The Biometric Mirror first takes a photo of the user. At this point, participants could be seen laughing, exclaiming in awe, or remaining cautiously still as the aesthetics of the “mirror” play before them. The captured photograph is then analysed by the system, applying facial analytics and algorithmic processes to generate a user profile with a series of statistics designed to incite further debate and discussion around the implementation of AI technologies.

The project’s art design is appropriately remnant of cyberpunk films from the 1980’s – blue neon colours cast against the stark white, procedurally generated statistical data. The results? A simple, but powerful, interface that lists human statistics as plain data, providing an algorithmically-generated profile to the user.

The first statistics that the Biometric Mirror is able to produce are those that have now become standardised details that consumers often give companies in a simple sign-up form – your age, gender and demographic information is generated, before a cautionary screen appears soon after. It asks the user, then, if they still want to proceed into the territory of more speculative data.

If the user proceeds, arguably subjective concepts such as agreeableness, weirdness, attractiveness, kindness and even introversion, are also rated by the computer – based solely on facial analytics. The accuracy estimates for these categories are transparently placed against the results – these could range from as low as 0%, to as high as a 99% accuracy rating. 

The project provided the ideal platform to spark a wide range of discussions and new areas for academic enquiry. From the ethics and psychology of facial analysis, to its efficacy as a model in determining abstract concepts about an individual’s personality – the study provides important launching pad for future AI research developments.

With pattern-recognition technologies already supporting many of our daily activities – smoother internet browsing through site cookies and cached data, advanced web search tools and location mapping systems – Biometric Mirror hopes to leave you with more questions than answers.

As AI technologies continue to develop, the team behind Biometric Mirror are encouraging us to continue to stay critical of the capacity for facial recognition technologies to operate independently from human intervention, as we advance further with the technologies we have today.

Would you believe what a computer had to say about you? Read more in Dr Woulter’s latest report, ‘Biometric Mirror: Probing the Ethics of AI and Facial Analysis’.