🔥 Đang khởi động dự án mã nguồn mở về giảm thiên lệch và cân bằng an toàn cho mô hình LLM. Mục tiêu: giảm bias mà vẫn giữ ổn định, an toàn. Mời mọi người quan tâm tham gia cộng đồng, đóng góp ý tưởng và mã nguồn. #AI #OpenSource #Debiasing #LLM #CôngNghệ #TríTuệNhânTạo #CộngĐồng

https://www.reddit.com/r/LocalLLaMA/comments/1qmpg6f/kickstarting_an_opensource_project_debiasing/

A 2021 paper in a pay-to-publish #Nature journal claimed #mindfulness tasks (notice 3 things around you or differences between 2 similar images) reduced some cognitive #biases: https://doi.org/10.1057/s41599-021-00712-1

The mindful #debiasing was undetected in a re-analysis and 2 replications: https://doi.org/10.1007/s12671-025-02575-y

Does awareness of intuition's fallibility help people avoid faulty intuitions?

#Teaching students #DualProcessTheory didn't help them avoid faulty intuitions about #physics problems.

https://doi.org/10.1103/PhysRevPhysEducRes.21.010135

#edu #cogSci #bias #debiasing #psychology #epistemology #rationality

More successful predictions of my #BoundedReflectivism & #EpistemicIdentity article from 2022?

From about 2017 or 2018, I reviewed evidence suggesting that reflective thinking often helps, but may even hinder our judgment depending on factors like whether we are reasoning based on shared identities and epistemic standards (https://doi.org/10.1111/meta.12534).

Years later Robb Willer et al. solicited ideas for debiasing political judgments. After years of testing 25 different interventions, their results are published in #Science: https://doi.org/10.1126/science.adh4764

They found that biased evaluations of politicized facts were impacted by both
- REFLECTION (on potential misperceptions of undemocratic actions) and
- shared IDENTITY cues (which could help *or* hinder!)

#decisionScience #politics #psychology #philosophy #epistemology #metaphilosophy #JDM #debiasing #chatBot

"Political #socialMedia ads labeled as from in-party sources and apparently non-partisan sources were trusted more than out-party sources.

So if you want to be sneaky just advertise under a name without any party connections"

That's a description from @mjbsp.bsky.social‬ about an #openAccess paper in #PoliticalPsychology by @tomstafford et al.: https://doi.org/10.1111/pops.13034

#politics #marketing #persuasion #journalism #decisionScience #nonPartisanship #socialPsychology #cogSci #bias #debiasing

This week's fifth and final post applies the reason-responsive #consequentialist view of rational inquiry to shed light on bounded #rationality, the Standard Picture, and the #epistemology of inquiry.

Thanks to #DavidThorstad for giving us a peak inside the book Inquiry under bounds (#OxfordUniversityPress 2024).

https://philosophyofbrains.com/2024/07/05/inquiry-under-bounds-part-5-applying-the-account.aspx

#philosophy #psychology #cogSci #decisionScience #economics #logic #philosophyOfScience #bias #debiasing

Does data visualization help people overcome mathematical biases like base rate neglect?

Not necessarily. In fact, one experiment on master’s students found that “when base rate information [was] presented visually, participants answered significantly more biased than when information [was] presented textually.“

https://doi.org/10.23947/2334-8496-2024-12-1-119-132

#probability #stats #edu #dataViz #bias #debiasing #argumentMapping

'High-Dimensional Inference for Generalized Linear Models with Hidden Confounding', by Jing Ouyang, Kean Ming Tan, Gongjun Xu.

http://jmlr.org/papers/v24/22-0834.html

#confounders #inferences #debiasing

Our pick of the week by Beatrice Savoldi: "Exploiting Biased Models to De-bias Text: A Gender-Fair Rewriting Model" Amrhein et al., 2023

https://arxiv.org/pdf/2305.11140.pdf

#NLProc #NLP #pickoftheweek #gender #bias #fair #fairness #debiasing #translation #MT

Calibrate and Debias Layer-wise Sampling for Graph Convolutional Networks

Yifan Chen, Tianning Xu, Dilek Hakkani-Tur, Di Jin, Yun Yang, Ruoqing Zhu

https://openreview.net/forum?id=JyKNuoZGux

#sampling #nodes #debiasing