Is that effect any different than the one you’d get if you have biased references, or biased search results, when doing the researchb for your writing?
Well of course it will be different. One has to do with another author publishing questionable data and the other would be related to misunderstanding of someone else’s published data. In this case, the use of AI in writing is implied to result in authors not being in control of what they themselves publish.
All of these are bad but do not necessarily arise on purpose. But let’s not add ways to muddy the already mudied waters of science.
Those seem like questions for more research.
I bet it’s more pernicious because it is easy to incorporate AI suggestions. If you do your own research, you may have to think a bit if the references/search results may be bad, and you still have to put the info in your own words so that you don’t offend the copyright gods. With the AI help, well, the spellings are good, the sentences are perfectly formed, the information is plausible, it’s probably not a straight-forward copy, why not just accept?
I’ve just read the abstract of the study - but it doesn’t seem to be about people mindlessly copying the AI and producing biased text as a result. Rather, it’s about people seeing the points the AI makes, thinking “Good point!” and adjusting their own opinion accordingly.
So it looks to me like it’s just the effect of where done view points get more exposure.
I am being brainwashed by AI!
Here’s the paper: https://dl.acm.org/doi/10.1145/3544548.3581196
Abstract
If large language models like GPT-3 preferably produce a particular point of view, they may influence people’s opinions on an unknown scale. This study investigates whether a language-model-powered writing assistant that generates some opinions more often than others impacts what users write – and what they think. In an online experiment, we asked participants (N=1,506) to write a post discussing whether social media is good for society. Treatment group participants used a language-model-powered writing assistant configured to argue that social media is good or bad for society. Participants then completed a social media attitude survey, and independent judges (N=500) evaluated the opinions expressed in their writing. Using the opinionated language model affected the opinions expressed in participants’ writing and shifted their opinions in the subsequent attitude survey. We discuss the wider implications of our results and argue that the opinions built into AI language technologies need to be monitored and engineered more carefully.
About what? Lmao
Maybe RTFA
