Companies need to be better at preventing racism and other forms of bias and discrimination from infiltrating their technology and systems. Consumer priorities are shifting and potential discrimination will no longer be tolerated. The implications of not flagging racially discriminative algorithms are huge and span all industries, because AI is now so extensively used. GlobalData forecasts suggest that the market for AI platforms will reach $52bn in 2024, up from $29bn in 2019.
Incidences of algorithmic bias are not new and have been covered in the media for over 10 years. As far back as 2010, image recognition software from Nikon cameras asked Asian customers if they were blinking and wanted to retake their photo.
These issues have been widely reported on, but shockingly they are still happening on a regular basis. Only last week the NYTimes reported that Facebook’s Image Recognition Systems labelled a video, uploaded by the Daily Mail, featuring black men, as ‘primates’. A caption beneath the video asked the viewer if they wanted to ‘keep seeing videos about primates?’
Facebook have since apologised and launched an investigation into why this unacceptable error occurred. Retrospective apologies are no longer enough and need to be replaced by more effective preventative measures.
How do bias incidences happen?
Bias can infiltrate algorithms in a number of ways. Many of these biases are related to the datasets which the algorithms are fed as part of their training. These algorithms learn bias through being fed data sets which are inadequate for their intended purpose.
Algorithms can be identified as inadequate if they are incomplete, not diverse enough or too small. Another common problem in image recognition systems is that the pictures in training datasets are frequently incorrectly labelled. These algorithms can then have varying effects on people of different nationalities, race and gender.
Changing consumer priorities
Consumer priorities have shifted since these issues first arose 10 years ago and now consumers expect higher ethical standards. GlobalData’s TrendSights Overview Report stated that there is a ‘growing distrust of companies, particularly multinationals’ as well as that ‘higher expectations drive consumers to be well-informed’.
Consumer education and extensive media coverage surrounding companies and their activities means that consumers are becoming ‘increasingly aware of the environmental and ethical issues that surround the products they consume’. This in turn is leading to a rise in ethical consumerism.
As a result, these incidents can be considered as not only ethically reprehensible, but also detrimental for a company’s future performance. It is easy to foresee the potential damage in terms of reputation and subsequent loss of sales for a company if these incidents continue. Damage to a company’s reputation has also never been so easy as it is now with the use of social media platforms and the ruthless nature of cancel culture.
It is disappointing that these incidents are still happening and that it might be consumer pressure that changes big tech behaviour rather than a more fundamental reassessment of corporate values.