Today's post is going to be a little different. It's an experiment in 'lean' writing. I'm going to compile a taxonomy of threats to democracy by AI, both near and long term... But I'm going to do it incrementally and iteratively. I'll start with some bullet points, and gradually update, sharing updates as I go. At first I'm not going to include references to existing work, but I'll gradually fill this in and gloss important work on each head of risk.
The list
Algorithmic curation of information / Socio-technical organisation of informtion production and consumption
prioritising engagement or clickthrough leads to unintended dynamics in the way information is prioritised
filter bubbles and echo chambers
facility of spread of misinformation, where misinformation too easily moves up the hierarchy of visibility
vulnerability of algorithmic curation to hacking by bots, including for the purposes of state-sponsored disinformation
outrageification and polarisation, including pushing people to extremes by failing to expose them to sufficiently diverse points of view
Targeted political and consumer advertising
State surveillance enhanced with facial recognition, voice recognition and data processing power
Location tracking by both state and non- state actors: threatens freedom of association
Automated decision-making: administrative decisions, law enforcement decisions etc, without sufficient explainability, accountability or opportunities for appeal
Availability of powerful and inexpensive autonomous weapons to non-state actors
Reactionary measures
For example, concern about 'harmful' or violent content achieving prominence online through unintended consequences of operation of curation algorithms leading to overreaching state-sanctioned application of algorithmic filtration and censorship
Comments