So, my point here was that "building public trust" only makes sense as a research goal in circumstances where such public trust is in fact a public good.
>>
>>
https://twitter.com/emilymbender/status/1487546391232536576
For example, public trust in effective vaccines (like the ones we have for COVID-19 and many other diseases) is important, because widespread vaccination brings about public goods (lower burden of disease, less pressure on the healthcare system).
>>
>>
Likewise, public trust in democratic systems (to the extent that they are truly democratic) brings about public goods, because broad voter participation makes for better governance.
>>
>>
But when we're talking about automated systems like large language models (and other #PSEUDOSCI aka #SALAMI), the key question isn't trust, but trustworthiness.
>>
>>
If someone tells you the point of their research is to build trust in AI/SALAMI/PSEUDOSCI they're selling you something ... or more likely, they're selling you (your data, your trust) to someone else, and it's not about public goods.
(For anyone reading this thread curious why SALAMI and PSEUDOSCI, see this tweet:
https://twitter.com/emilymbender/status/1485067239023775744
• • •
Missing some Tweet in this thread? You can try to
force a refresh
