Estimate the number of annual excess deaths globally that are causally linked to the deployment of engagement-maximizing algorithms (YouTube, TikTok, Facebook, X, etc.).
Counterfactual: How many fewer people would die this year if content feeds were not engagement based? The total stack of externalities, ranging from direct violence to indirect killers like lifestyle diseases.
Case for a high estimate:
Sedentariness: The WHO attributes ~3.2 to 5 million deaths annually to physical inactivity (linked to heart disease, diabetes, stroke). e.g. An increase by 1-10% could mean 30k-500k deaths/year.
Genocide and war: Algorithms prioritize engagement, emotion (rage/fear) is very engaging. Side effect e.g. Facebook role in Rohingya genocide (Myanmar), estimation for Tigray War (Ethiopia), Syria, Sudan, Nigeria, India..
+ If it cause polarisation, increasing probability of war.
Mental health: With billions of users, even a 0.01% increase in suicide rates or fatal eating disorders attributable to AI nudges results in tens of thousands of excess deaths.
Disinformation: e.g. medical disinformation. e.g. if it increase % of 100k measle case/year, of 20M saved live by COVID vaccine in 1st year, etc
1,000Additional thought:
Estimation of the amount of hour of content watched by human everyday on YouTube + Facebook + Instagram + Tiktok?
What if those recommendations AI where aiming at "benefiting society" (recommending information on topic collectively identified as important, interesting etc) instead of "maximising attention"?
How much money and time would it be worth investing in this problem?
How would this size effect compare to investing money and time in others important society problems?