Eve Fleisig (@enfleisig) / X

Eve Fleisig

242 posts

Eve Fleisig

@enfleisig

Incoming @PrincetonCITP fellow | PhD @Berkeley_EECS | Princeton ‘21 | NLP, AI ethics + equity, and sociolinguistics enthusiast | bilingüe 🇦🇷

Joined July 2020

Pinned
Eve Fleisig
@enfleisig
Jun 17
Remember the backlash against all AI voices being feminine—Siri, Alexa, etc.? Now, companies use a wider range of AI voices with ambiguous names like “alloy” and “shimmer.” Bias solved forever—ha, no. What happens when people hear these supposedly neutral voices?👇
19K
Eve Fleisig
@enfleisig
Dec 12, 2023
🎉Our work has received the Outstanding Paper Award in the Ethics in NLP track at #EMNLP2023! We're so honored @emnlpmeeting😊
Eve Fleisig
@enfleisig
Dec 7, 2023
Has collecting data on MTurk ever felt mystifying? To synthesize best practices, our #EMNLP paper studies MTurk worker perspectives on issues like data quality, privacy, fair pay & worker incentives. Surveying workers led to some surprises… 📄arxiv.org/abs/2311.02802 🧵👇
13K
Eve Fleisig
@enfleisig
Jun 18, 2024
What should we do when annotators disagree? Our position paper (w/ @ZeerakTalat, @sulin_blodgett, Dan Klein) charts issues faced by the longstanding data collection paradigm & new *perspectivist* approaches that treat disagreement as signal, not noise. 📄 arxiv.org/pdf/2405.05860
7.7K
Eve Fleisig
@enfleisig
Jul 9, 2023
Can we better capture harms in text generation? Our #ACL2023 paper introduces FairPrism, a dataset with detailed human annotations on AI-generated text to diagnose fairness-related harms caused by text generation systems & flaws of mitigation tools🧵 📄bit.ly/fairprism
18K
Eve Fleisig
@enfleisig
May 24, 2023
Aggregating data labels for tasks like hate speech detection masks real differences in opinion. Predicting individual annotators' labels lets us model the opinions of those harmed & catch high-uncertainty cases while still accounting for privacy... 📄 arxiv.org/abs/2305.06626 🧵👇
14K
Eve Fleisig
@enfleisig
May 28, 2024
I’ll be at Stanford on Thursday, for anyone who’s interested in chatting!
Stanford NLP Group
@stanfordnlp
May 28, 2024
For this week’s NLP Seminar, we are thrilled to host @enfleisig to talk about "Linguistic Bias in ChatGPT: Language Models Consistently Reinforce Dialect Discrimination"! When: 05/30 Thurs 11am PT Non-Stanford affiliates registration form (closed at 9am PT on the talk day):
9.2K
Eve Fleisig
@enfleisig
Apr 24, 2024
NAACL 2024 seems to charge $750 for students to register if they're a presenter (every paper requires at least one registered presenter). @naacl am I reading this right? Seems like a major burden on students, especially if (as is common) only a paper's student authors attend.
25K
Eve Fleisig
@enfleisig
Dec 7, 2023
Has collecting data on MTurk ever felt mystifying? To synthesize best practices, our #EMNLP paper studies MTurk worker perspectives on issues like data quality, privacy, fair pay & worker incentives. Surveying workers led to some surprises… 📄arxiv.org/abs/2311.02802 🧵👇
arxiv.org
Incorporating Worker Perspectives into MTurk Annotation Practices for NLP
Current practices regarding data collection for natural language processing on Amazon Mechanical Turk (MTurk) often rely on a combination of studies on data quality and heuristics shared among NLP...
17K
Eve Fleisig
@enfleisig
Dec 7, 2023
I’m at EMNLP this week! Come say hi if you’re interested in… How we can make use of annotator disagreement: (11-12:30 Fri) MTurk best practices that incorporate worker perspectives (2-3:30 Fri) Operationalizing marginalization with outlier-based methods (11-12:30 Sat) Papers👇
4.8K
Eve Fleisig
@enfleisig
May 6, 2024
📄🚨 Social choice theory often comes up as a reference point for work on RLHF. @jessicadai_ and I went on a social choice deep dive to investigate how we can (and can't) map these problems together. 📄 arxiv.org/pdf/2404.13038… 🧵
5.1K
Eve Fleisig
@enfleisig
May 7, 2024
I'm at ICLR this week! Come say hi if you're interested in: - Fairness, equity, & ethics in NLP - Annotator disagreement/perspectivism - Social choice theory + RLHF
3.2K
Eve Fleisig
@enfleisig
Jul 8, 2024
Thanks to the people who generously volunteered over the past few days, we now have enough emergency reviewers. Much obliged to everyone who signed up! 💕
ACL SRW 2025
@acl_srw
Jun 22, 2024
We're seeking additional emergency reviewers: please fill out the form below if you're available to support the workshop!
3K
Eve Fleisig
@enfleisig
Jul 6, 2024
ACL SRW is a great opportunity for students starting out in NLP. Unfortunately, it’s been sadly common for reviewers not to submit their SRW reviews and the SRW needs more emergency reviewers in the next few days. You can help us out by signing up here: forms.gle/5b5FNAS7oCCZ5a…
ACL SRW 2025
@acl_srw
Jun 22, 2024
We're seeking additional emergency reviewers: please fill out the form below if you're available to support the workshop!
14K
Eve Fleisig
@enfleisig
Jun 17, 2024
We’ll be presenting on this at #NAACL2024 today at 11–come stop by our poster!
Eve Fleisig
@enfleisig
Nov 14, 2023
📄🚨Check out our new work (with @vcubingx @NickATomlin and Dan Klein) on Ghostbuster, a new SOTA method for LLM-generated text detection!
1.8K