Question to Isabelle Falque-Pierrotin :
According to professor Karen Yeung, an algorithm-based decision is acceptable as long as it can effectively be contested in court. In a context of a more heavy algorithmic regulation on online plateforms, decisions such as the one taken by la Cour d’appel de Paris on the 12th of october show the relevance of national juridictions to deal with such cases. Yet, specialised institutions such as yours seem to be overcrowded with demand, as pointed out in the LFI 2017.
So, are the french and european juridictions able to offer an effective recourse right against online algorithm based decision?
- Yeung (Karen), “Algorithmic regulation: A Critical interrogation“
- How social media filter bubbles and algorithms influence the election https://www.theguardian.com/technology/2017/may/22/social-media-election-facebook-filter-bubble(for an example of facebook crackdown on alt-right groups)
Thank you for this interesting question. This exactly reminds me of another moral dilemma that I discussed with my classmate today. Maybe it’s a bit off topic, but I think that it still adds some spice to the discussion. As programmers need to conceive algorithms for autonomous vehicles, they should think of the actual application of the famous moral question: when you go ahead, you will crash four people; when you shift in another direction, you will crash one person, how will you choose? So when algorithms need to be encountered with moral issues, how can programmers decide? And should these decisions be justifiable?
Good question ! The CNIL is nowadays striving to implement an European juridiction regarding the data privacy, and therefore focus on the sourcing of the algorithm (data gathered by the algorithm to be more efficient). The debate seems to be focus on this issue, forgetting about the use companies and public institutions make of algorithms. In the United States, there are used for many decisions, that can have terrible impacts on people lives : in Court (crime scores), in banks and insurance companies (FICO score), in schools (school registration). These algorithms are considered as purely objective, as they are mathematical – and people rely on them so that they will not be accountable for their decision (ex: a judge in Court condemns to 10 year of prison a guy with a high crime score, and justify its decision by saying that he relied on the result of the algorithm). How can we juridically contests such decisions?