I'd suggest that we figure out what our algos are measuring, first. Given how new this field is, and how many different disciplines are looking at it, we need

I'd suggest that we figure out what our algos are measuring, first. Given how new this field is, and how many different disciplines are looking at it, we need a common set to work on. "Online Harms", was the broadest set of terms I found. (UK Gov - https://www.gov.uk/government/consultations/online-harms-white-paper https://www.gov.uk/government/consultations/online-harms-whi...) Here is count of categories in transparency reports for 3 firms (full list at the end): * Online harms paper: 24/22 harms * Facebook, 10 harms (1) * Twitter 12 (2) * Reddit 8 (3) Here's a paper from May 2020, which focused on hate speech detection using SA. The intro lays out the case for clearer, more compatible categories: "the lack of clear definitions of key categories is a critical issue. The authors argue that researchers use different, sometimes theoretically ambiguous or misleading terms for equivalent categories." https://www.aclweb.org/anthology/2020.lrec-1.838/ https://www.aclweb.org/anthology/2020.lrec-1.838/ Which brings us to the problem of secrecy in tech firms. ------ What we know about how the firms work, is based on leaks and people who break NDAs. Reasons range from: 1) It's a trade secret 2) An open list means attackers will use new words 3) It exposes their approach to criticism. To whatever. Researchers are likely recreating the wheel. The question is whether the information on the "public square" needs to be released and made open. This punts the question on regulating algos, to after we figure out what algos are working on in the first place. After that, maybe you can start talking about algorithms which curate, ban, promote, profile anything that humans do. --------------------------- (1) FB harms: Adult Nudity and Sexual Activity, Bullying and Harassment, Child Nudity and Sexual Exploitation of Children, Dangerous Organizations: Terrorism and Organized Hate, Fake Accounts, Hate Speech, Regulated Goods: Drugs and Firearms, Spam, Suicide and Self-Injury, Violent and Graphic Con (2) Twitter harms: Hateful conduct, Abuse/harassment, Sensitive media, Promoting suicide or self-harm, Violence, Illegal or certain regulated goods or services, Non-consensual nudity, Private information, Impersonation, Child sexual exploitation, Civic integrity, Terrorism/violent extremism (3) Harassment, Minor Sexualization, Violent Content, Involuntary Porn, Controlled Goods, Private Information, Impersonation, Ban Evasion