Select a result to preview

discrimination

Discrimination

Definition of discrimination

Discrimination in AI specifically can be defined as objectionable or illegal discrimination based on for instance gender, skin color or racial origin that is caused by AI. It is often related to machine learning algorithms and the bias that exists in the data that is fed to these algorithms. The difference between bias and discrimination is that bias is the tendency of an algorithm to be skewed based on input data, and discrimination happens when this bias actually results in the unfair treatment of certain groups based on non-relevant features.

Implications of commitment to discrimination

If one commits oneself to diminishing discrimination by AI, one should first and foremost be fully aware of the possible causes of this discrimination in order to prevent it. The most important part of designing an algorithm that should be focussed on when trying to achieve this goal is the input and training data that is used to train an algorithm. Like I said earlier, discrimination by AI is often the result from bias in training data. It would therefore be most important to ensure that the training data does not contain any bias, or features that could cause discrimination. In this process, it is very important to keep in mind that discrimination might result from many different features that one would not think about at first hand. It is therefore important to eliminate any proxies that might lead to discrimination as well.

Societal transformations required for addressing concern raised by discrimination

Discrimination is something that unfortunately is still happening in all layers and institutions of society. I think that is therefore of the utmost importance that discrimination is an issue that is recognized and addressed globally in order to make this movement in AI really happen. Even in our own educational system, we still are seeing changes being made to curricula because they included discrimination or certain ethnical groups are severely underrepresented in the education.