Now that we have familiarised ourselves with the definitions and regulations within the field of artificial intelligence and automated decision-making, it is time to take a look at the areas where these technologies are used and the potential risks they pose within the field of anti-discrimination.

 

 

 

Please take the time to familiarise yourself with the relevant pages of the policy brief Preventing the potential discriminatory effects of the use of artificial intelligence in local services.

Food for thought

Take some minutes to consider the following questions which may help you identify areas to address:

  • What can your city do to ensure the automated decision-making systems have as complete data as possible to base the decision-making on?
  • Can you think of practical situations where the use of technological tools in your city has led to, or could have led to, discriminatory decisions? Why have these situations occurred? What could your city have done to stop this from happening?
WANT TO LEARN MORE?
Knowledge test

Continue »

6. What is statistical discrimination?

  • (1)Unjustified unequal treatment of persons on the basis of surrogate information
  • Using statistics to help make decisions
  • Unequal treatment based on the personal, prejudiced dislikes or preferences of the decision-makers against a certain group of people or on dislikes or preferences for certain products

Statistical discrimination is the unjustified unequal treatment of persons on the basis of surrogate information. However, it is crucial to understand that, as algorithms are created by humans with all their biases included, statistical discrimination can originate from the taste-based discrimination. These two phenomena are therefore very rarely independent of each other.

7. How can automated decision-making result in discrimination?

  • (1)We have incomplete information about the person
  • The AI does not carry out the task as instructed
  • The automated decision-making does not include a human decision
  • (1)The training set may not be representative or correctly represent society

Automated decision-making often fails to include an extensive evaluation of the circumstances of the individual and is dependent on the information, data, and settings it has been trained in. Context and full information are crucial to avoid unwillingly biased decisions.

Knowledge test results

[[score]]

 

↻ Start over Go to module 3 »