Data Has As Much Potential For Evil As Good

It has as much potential to do good than bad


Here at the channels we believe that data is an overall force for good. Having spent years reading and writing around the subject, we have seen it being used to protect forests, work to cure cancer, and improve people’s lives around the world.

However, data can be used for as much bad as good, and a story last week really highlighted this issue.

An AI algorithm created in a study from Stanford found that it could distinguish the sexuality of men and women from their photos alone. It was accurate in 81% of cases for men and 74% for women from a bank of 35,000 images posted on US dating sites. Once it saw 5 images this number actually increased to 91% accuracy for men. It also found some startling trends, as The Guardian reported:

‘The research found that gay men and women tended to have ‘gender-atypical’ features, expressions and ‘grooming styles’, essentially meaning gay men appeared more feminine and vice versa. The data also identified certain trends, including that gay men had narrower jaws, longer noses and larger foreheads than straight men, and that gay women had larger jaws and smaller foreheads compared to straight women.’

This kind of use of data and data technology is largely harmless when it’s in the confines of a relatively benign academic study, but if this kind of technology is being developed, it creates the potential for it being used for considerably more sinister things.

It is something that Michal Kosinski, co-author of the study and an assistant professor at Stanford, is well aware of. He says he was highlighting how this technology could potentially be used for evil, stating: ‘One of my obligations as a scientist is that, if I know something that can potentially protect people from falling prey to such risks, I should publish it.’

There are several regimes around the world where being gay is illegal or actively persecuted. For instance, in Chechnya there has been widespread reports of homosexual people being tortured and killed, whilst it is still illegal to be gay in Iran, Saudi Arabia, India, Egypt, and over 70 other countries. If this kind of data technology were to be used by these governments to identify and persecute gay people it could be catastrophic and in some cases could even lead to active genocide.

Unfortunately, data and data technologies are cold, unfeeling, and calculated, looking at only the facts they have, rather than looking at potential discrimination or systemic bias that created the data from which they operate. For instance, in the US, black people are 3.6 times more likely to be arrested for selling drugs, whilst white people are 45% more likely to sell drugs. If the arrest rates were fed into a recruitment algorithm this data would tell the AI that its better to not hire black people as they are more likely to have committed a crime, which is simply not the case despite what the data tells the system.

In a similar vein, the use of data that was collected and held in confidence can potentially be used for more sinister means in the wrong hands. Although it may not get to that stage, there is the potential for it when Donald Trump ended The Deferred Action for Childhood Arrivals (DACA) about which he said on the campaign trail ‘We will immediately terminate President Obama’s two illegal executive amnesties, in which he defied federal law and the constitution to give amnesty to approximately 5 million illegal immigrants.’ The basis of DACA was that through registering themselves with extensive records, including medical records, school records, military records etc, people who were brought to the US as children could remain and work in the country with no danger of deportation. If the program does end, this data which was given to the US government in good faith could easily be used as a tool to remove these people.

The last decade has taught us that anything can happen and change the way we thought established society worked. At present we are collecting a huge amount of data and if that is used in an incorrect way, it could have far reaching consequences. 

Looking small

Read next:

Expert Insight: 'An Effective Visualization Results From A Great Deal Of Curiosity And Exploration'