The Way People Judge Your Appearance Could Change Your Future
Is our future ruled by algorithms and artificial intelligence?
We know we're not supposed to judge others based solely on appearance. The reverse is frustrating too -- when someone takes one look at you and forms an opinion.
Most of the time we can shrug off the judgment or hear our inner voice monitoring our thoughts with the 'don't judge a book by its cover' principle.
The presence of artificial intelligence (AI) however, poses the threat that judging people based on their appearance could become an institutionalised practice. Problems could arise should the government, or your insurer, a future employer or a teacher get their hands on information that deems you irresponsible, weird or unhappy based on what other people think of your appearance.
"Very often AI comes with a consequence ... [when] AI systems assess someone to be quite irresponsible ... automatically that would make that person illegible for management positions," Dr Niels Wouters from the University of Melbourne told ten daily.
Dr Wouters and his team has developed the Biometric Mirror to demonstrate some worrying trends with AI. The Mirror shows research participants how they are perceived when they walk down the street based on a single snapshot of their face.
The Mirror was developed with data collected from a group of people all over the globe who evaluated 10,000 photos of others. Eighty-five percent of this group were caucasian individuals. Each photo was assessed based on 14 characteristics, including attractiveness, emotional stability, weirdness, age, gender and ethnicity.
"We've fed that into that artificial intelligence model and that basically allows us to automate the process and to upload a photo of a person and get an idea of what these people might think of that person," Wouters said.
"It is purely based upon subjective feedback of people."
But it seems flawed to judge people based on how they look right? Well, that is exactly the point of Wouters' research.
As it stands, AI relies almost entirely on a data set based on highly subjective feedback from people in society. Data bias is influenced by a number of factors, including culture, ethnicity, place of residence and societal environment.
"To make the ultimate unbiased artificial intelligence system, or at least the unbiased facial recognition system, we need photos from 7 billion people basically ... that's from every person on earth," Wouters said.
Wouters' team cannot survey every person on the plant so, yes, the results of their research will be flawed due to such a small data set.
But that's precisely the point -- to emphasise on a smaller scale the ethical repercussions artificial intelligence could have if it's used irresponsibly.
The Mirror's assessment of some of ten daily's staff demonstrated this.
The flaws found in the six examples sent by ten daily to Wouters were widespread. The discrepancies ranged from age and ethnicity to how aggressive people were or how happy they deemed themselves to be.
For example, Myles Davies is 47 years old, not 67, and Mat Whitehead is of Caucasian ethnicity, not African.
"We ... want to show the scenarios to people and make the point that AI can be massively flawed," Wouters told ten daily.
"The discussion is more, 'Is that really a direction that we want to go into as a society? Do we really want to rely on algorithms to look at faces and make decisions based on face?
"Someone who might be perceived as not attractive or irresponsible by an algorithm, that doesn't necessarily mean that that person is unsuitable for a job or is not nice, is not sociable, so these are the discussions that we need and we urgently need to have," Wouters said.
Ultimately, Wouters and his team are looking for public debate and awareness about the world of emerging technology.
He called for everyone from technology developers to academics and the general public to be involved in such a discussion because the consequences of AI being misunderstood or used incorrectly on a large scale could affect everyone.
The urgent need is policy debate.
"We need to develop policy around artificial intelligence for decision making and that could manifest itself in various forms," Wouters said.
"Maybe we need to be more transparent when it comes to computers and systems and advertise what the data is being used for, what characteristics are distinguished and what the consequences might be."