That Word ChatToday at 4:30 ET!
Experimental psychologist Corinne Jorgenson is back on That Word Chat to explore how AI is reshaping the editor’s role, what “human-aligned” really means, and more.
Watch the YouTube Premiere here: bit.ly/ThatWordVideo
Huge thanks to Corinne Jorgenson for joining us on #ThatWordChat today. From cognition to ChatGPT to ethics, this conversation gave us so much to think about.
That means removing identity indicators like gendered language, names, and zip codes before training the model. This is one approach to reducing systemic bias in output. -Jorgenson #ThatWordChat
As AI outputs more biased content, it risks amplifying the same voices over and over, “moving that bias needle even further in the wrong direction...There are hyperparameters now that help tune out some of that bias, but it’s not perfect." #ThatWordChat
One major concern is bias. “These tools were trained on data that’s mostly English, mostly published, and mostly reflective of a narrow set of voices: certain speech, certain status, certain skin color.” -Jorgenson #ThatWordChat
ChatGPT now includes memory, “going through the history of all of your chats… your specific syntax, the way you write.” It can even generate content in your voice. “Super cool, also super dangerous.” -Jorgenson #ThatWordChat
Jorgenson is studying how humans interact with systems like ChatGPT and whether being able to explain why the system responds a certain way builds trust.
“The idea is that explainability could create trust, and trust enables better cooperation.” #ThatWordChat
“The idea is that explainability could create trust, and trust enables better cooperation.” #ThatWordChat
Readers recognize AI-generated content now.
Those familiar phrasings—“In a world where…” “In today’s technological landscape…”—trigger distrust. “If you feel like you can’t credibly claim the work as your own, that’s when you should probably disclose.” #ThatWordChat
Those familiar phrasings—“In a world where…” “In today’s technological landscape…”—trigger distrust. “If you feel like you can’t credibly claim the work as your own, that’s when you should probably disclose.” #ThatWordChat
A more meaningful question, she says, came from another proposal: Did you use AI to generate your ideas or deepen your understanding of the solution? "Not whether you touched a tool, but whether it shaped the thinking behind your work."-Jorgenson #ThatWordChat
In one recent research proposal, Jorgenson saw a disclosure request: Had AI been used in the submission at all?
“In today’s world, if you used a computer, you probably used generative AI, whether you realize it or not.” -Jorgenson #ThatWordChat
“In today’s world, if you used a computer, you probably used generative AI, whether you realize it or not.” -Jorgenson #ThatWordChat