Scientists have used algorithms to identify dating site users as gay or straight – but what would this be used for?
In the battle of man vs. machine, machine may now have a stronger “gaydar” than us. According to a new study, it is now possible for technology to accurately predict – with a high score – whether a person is gay or straight, just by analysing a photograph.
Stanford University has tested AI’s potential to accurately label a person’s sexual orientation by using an algorithm on around 35,000 images . It was successful 81 per cent of the time analysing gay men and 74% when analysing an image of gay women. Dr Michal Kosinski revealed that the algorithm outperformed human beings taking the same test.
The focus of this tech includes both permanent facial features and less fixed areas, i.e. your grooming style. This means that based on the size of your forehead and the length of your nose, the machine is able to make a prediction. This feature of the machine adheres to the prenatal hormone theory of sexual orientation, which suggested gay men and women tended to have gender-atypical expression and grooming styles.
This comes from the idea that during the development process, fetuses are exposed to various levels of hormones, particularly testosterone. The researchers state that such hormones assist in developing the facial structures of the fetus which can be involved in determining sexuality, strengthening support for the argument that people are born gay, rather than choosing to be.
These findings have come as a result of ever-growing volumes of data and improving algorithms which highlight the increasing power of technology and its potential to seek out and inevitably harm individuals.
Dr. Michal Kosinski of Stanford University has collaborated with Yilun Wang in conducting this project. Kosinski has previously released works on Facebook data and its use for detecting personality traits. Kosinski also mentions that with the right combination of data, similar systems may be able to spot other traits such as IQ or political views, which is also pretty terrifying.
The study has been met with criticism. In a statement, Ashland Johnson of the Human Rights Campaign (HRC) said: “Imagine for a moment the potential consequences if this flawed research were used to support a brutal regime’s efforts to identify and/or persecute people they believed to be gay”. Indeed, recent events in Chechnya demonstrate the violent persecution and discrimination that gay people still face around the world.
“Imagine for a moment the potential consequences if this flawed research were used to support a brutal regime’s efforts to identify and/or persecute people they believed to be gay” – HRC
In today’s heternormative society, coming out remains a personal decision and surely not something that should be guessed by another human being controlling a machine?
Regardless, the technology has a long list of limitations. Firstly, it is not proven to work on all photos or all types of people – trans people and bi people are not accounted for. Bizarrely, no people of colour were included in the study. The software doesn’t seem to be able to accurately predict for women as it can for men.
Kosinski revealed that the prediction models with the sole aim of gender resulted in less accurate predictions. It detected gay males with 57% accuracy and gay females with 58% accuracy, contradicting a previous idea that women were more fluid in their orientation and less easy to detect.
“There are still areas across the globe where being gay is socially unacceptable, or even illegal”
This project is fairly small scale and thus far only tested on one US dating site. However, there are still areas across the globe where being gay is socially unacceptable, or even illegal, which raises concern. In the future, this tech could be severely dangerous or threatening to the safety – both online and in the real world – of gay men and women. The software could be used by corporations and governments to target ads or attacks. Understandably, the study has caused a surge in anger, upset and worry within the gay community.
While scientifically impressive, it doesn’t seem morally correct for this tech to exist in our current world, as the science places members of the gay community in a position of vulnerability – it seems likely that man would only use this machine for harm.