Are algorithms racist?

Last week in our newsletter, we discussed Chat GPT, a chat system based on the GPT-3 AI language model developed by OpenAI . The arrival of this new tool has reignited an old debate about AI and its potential for racism, misogyny, or sexism.

Artificial intelligence is here to stay and its impact on our stories may be unprecedented, so we must start raising these kinds of debates before it's too late.

“ Government policy is becoming increasingly ineffective in developing large-scale solutions to address the problems of bias and racial discrimination in algorithms, ” according to Public Citizen .

This assertion could pose a problem given that businesses and governments have increasingly relied on algorithms for decision-making. Their use is becoming increasingly widespread among nations for identifying potentially criminal profiles or guiding medical decisions that affect millions of people.

We tend to think that few things in this world are more reliable, neutral, and less biased than a good algorithm. But this can be a misconception. A study by Harvard and Georgetown shows that the algorithms we use can discriminate against certain races, genders, or social groups. Historically disadvantaged groups, such as women and Black people, suffer the most from algorithmic bias.

Researchers investigated the true cause of this phenomenon, eventually realizing that the algorithms were learning from and adapting to the inequalities that existed in our environment. Algorithms can reflect and perpetuate discrimination and prejudice present in society and in the data used to train them. If the data used to train an algorithm is biased or prejudiced, the algorithm itself may also be biased or prejudiced.

So, is the problem with the algorithms or with the data that feeds the algorithm? We leave you with this question to reflect on and share in the comments of this post.