Get Badass about Bias in AI Campaign
What is unconscious bias?
Using Facebook as an example, they use algorithms, to determine what to show you in your news feed from all the tons of information that is out there. What is ultimately shown to you is based on the past information you have shown an interest in. Unconscious bias is much the same, in any given moment your brain is bombarded with 11 million of pieces of information. Consciously, however, you can only be aware of 40 bits. This means, there were tons of vital information about a given situation that you missed. Without much thought, therefore, we make snap judgements about people's race, gender, sexual orientation, religion or nationality, especially when we are tired or rushed. Our biases impact how we see the world, how we treat people and the decisions we make. We are all biased by the way. We wouldn't be human if we wasn't. The conclusions we come to are influenced by our backgrounds, environment, personal experiences and stereotypes. Awareness of humans unconscious bias was increased with the Implicit Association Test. Check out our resources page for more educational videos on the topic.
What is Artificial Intelligence (AI)?
Computer systems that act and think like humans. Like humans too, they can translate languages, recognise voices, recognise faces and from reading text can decide if you are happy, sad or even in love. For example, just like a family member can recognise your voice when you speak to them over the phone, it is artificial intelligence that can recognise your voice when you speak into your phone. You'll hear the term machine learning banded around quite a bit too. This is the arm within AI that is creating the most excitement at the moment. It's where machines have literally been coded to learn just like how humans learn. Give it information (data) and just like a human, it can learn from its mistakes until it comes up with the right strategy. It just does it a hell of a lot quicker than we do and with less psychological issues :)
Sounds great to me and so what's the problem?
Indeed, like humans, the answers the computer provides will be based on the data it has been fed. The problem is, the data is fed to it by humans who have deeply ingrained unconscious biases which the computer inevitably adopts. Currently, humans can at least be made aware, counteract and reconsider any stupid conclusions it may have come to. AI is unable to do the same.
Ah, I see, but this will take a long time to catch on, no?
Organisations in every sector from banking to healthcare are investing trillions in AI to automate their business processes. In truth, you can't blame them. It is fast, efficient, reliable and saves them tons of money in the long run. The investment by organisations into AI is on the increase and predicted to jump from 54% to 63% in three years time. (Source: 2017 Global Digital Survey).
In a nutshell, your problem is?
The rapid rise in the use of AI means prejudices will become reinforced, especially those damaging to women, young girls and minority groups.
How about Government intervention, a regulator so to speak to ensure fairness and transparency?
Yes, experts are on this, however, due to the complexity of AI, they have all admitted to the struggle in getting it regulated. Even new laws coming in place, experts argue, simply do not go far enough. Check out this article.
And what does the Campaign want?
Due to the problems in getting AI regulated, it is clear that the change needs to start with us. We are campaigning for the UK government to help strip the jargon from what is a complex area and to launch a major awareness campaign to raise awareness of unconscious biases in AI and provide help to override it.
Please join the campaign.
Statistics and facts:
- Research supports the idea that if you stop and give yourself just a moment to think you can make the unconscious conscious.
- A recent study shows girls and young women feel held back by gender stereotypes, sexism, and anxiety about how they look.
- There was also a significant increase this year of mental illness among young girls due to increase pressure on how they look.
- Mashable reported on a study which showed girls were affected by stereotypes by age six.
- AI is on the increase and predicted to jump from 54% to 63% in three years time. (Source: 2017 Global Digital Survey).
- Research published in the journal Science demonstrates how biases are replicated in machine learning tools.
- The Gartner market research company forecasts that the era of smart machines will become one of the most disruptive phases in the history of IT (Source:Siemens.com)
- Companies in the top quartile for racial and ethnic diversity are 35 percent more likely to have financial returns above their respective national industry medians. (Source: Mckinsey)
- Companies in the top quartile for gender diversity are 15 percent more likely to have financial returns above their respective national industry medians (Source: Mckinsey)
- Recent research from Robert Walters, supported by the Employers Network for Equality & Inclusion (enei), has found that while 85% of employers consider building a diverse workforce to be a priority, a staggering 46% have zero strategies or programmes in place to achieve it.
- And a 2012 Deloitte study captured the views and experiences of 1,550 employees at three large Australian businesses operating in manufacturing, retail and healthcare, and identified an 80% improvement in business performance when levels of diversity and inclusion were high.
- The New York Times highlighted a study by Yale University which concluded that "Science professors at American universities widely regard female undergraduates as less competent than male students with the same accomplishments and skills."
- This Guardian article and the Science highlights how just like humans Artificial intelligence can be both sexist and racist. Joanna Bryson, a computer scientist at the University of Bath and a co-author said: “A lot of people are saying this is showing that AI is prejudiced. No. This is showing we’re prejudiced and that AI is learning it."
- An article on the BBC's news website highlighted the work done by Joy Buolamwini, a postgraduate student at the Massachusetts Institute of Technology. She was unable to carry out her work on face recognition algorithms unless she wore a white mask as it did not recognize her dark skin.
- New York Times article on a study that raised concerns about high paying jobs not being shown to women in search results and data that resulted in over-policing in predominately black areas because of software that was biased against blacks.