Quiz · Algorithmic Bias Debate

Question 1What is the main purpose of the students' discussion?

To catalogue every known failure of automated decision systems To prove that algorithms can never judge people fairly To prepare a joint seminar paper on bias in automated decision-making To design a new recruitment tool for technology companies

Question 2What did further analysis reveal about the biased recruitment tool?

Its programmers had deliberately designed it to discriminate It favoured female candidates for technical roles It had been built with no training data at all The problem lay in biased historical hiring data

Question 3Why did the healthcare algorithm underestimate the needs of poorer patients?

It was designed with malicious intent by the hospitals It used past spending as a misleading measure of illness It ignored the spending records of all the hospitals It relied on data collected from a single hospital

Question 4What did Rachel's example of the recruitment tool show?

It had systematically favoured male candidates over equally qualified women It had rejected all candidates for technical roles automatically It had been deliberately programmed to ignore female applicants It had improved fairness across a decade of hiring decisions

Question 5Whom did the company initially blame for the recruitment tool's bias?

Its programmers, before analysis pointed to the historical data The female applicants, for providing incomplete information The hospitals, for sharing misleading spending records The government, for failing to regulate automated systems

Question 6What was Karim's objection to blaming the data for biased outcomes?

It ignored the choices made by the teams who built and deployed the systems It placed too much responsibility on individual programmers It assumed the historical data could never be corrected It overlooked the spending records used by hospitals

Question 7Which statement is closest in meaning to Karim's view of neutrality?

Neutrality can be achieved if training data is cleaned properly Neutrality is an illusion because design decisions embed assumptions Neutrality matters less than the speed of the algorithm Neutrality is guaranteed when teams follow company rules

Question 8What conclusion did the group reach about improving automated systems?

Perfect algorithms should be demanded from the very start Programmers should be held legally responsible for all bias Historical data should be deleted from all training sets Auditing systems regularly matters more than demanding perfection

Question 9How did the students decide to structure their seminar paper?

Around remedies rather than simply cataloguing failures Around a history of recruitment tools in technology companies Around the programmers blamed by the companies involved Around comparisons with earlier essays on healthcare

Question 10From which field did Sofia take her concrete example of algorithmic bias?

Recruitment, based on a decade of hiring decisions Online retail, based on customer spending patterns Healthcare, where an algorithm underestimated poorer patients' needs Banking, where an algorithm refused loans unfairly