Algorithmic accountability
mercredi 30 décembre 2020 à 01:00The "second wave" of studies on algorithmic accountability is taking up questions about which jobs algorithms should be used to do.
I think the crucial underlying question about whether to job algorithmically is whether to collect the data needed to do it algorithmically.
Collecting data about people in a database is inherently threatening and dangerous — more so for some data than for others. If the algorithm requires dangerous data collection, that is a reason not to use the algorithm.
It is a mistake to assume that the danger affects mainly "marginalized communities". Many dissidents come from marginalized communities, because they campaign for justice for those communities. Whistleblowers, by contrast, usually have to be insiders in order to come across the secrets they reveal. And we all depend on them, whether our communities are marginalized or not, to give us the opportunity to restrain our countries from crime and oppression.