CINXE.COM

Maarten Sap - Home

<!DOCTYPE html> <html lang="en"> <head> <link rel="stylesheet" type="text/css" href="tools/style.css"> <!-- Google tag (gtag.js) --> <script async src="https://www.googletagmanager.com/gtag/js?id=G-T8EHFV852B"></script> <script> window.dataLayer = window.dataLayer || []; function gtag(){dataLayer.push(arguments);} gtag('js', new Date()); gtag('config', 'G-T8EHFV852B'); </script> <meta charset="utf-8"> <meta http-equiv="X-UA-Compatible" content="IE=edge"> <meta name="viewport" content="width=device-width, initial-scale=1"> <meta name="description" content=""> <meta name="author" content="Maarten Sap"> <link rel="stylesheet" href="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/css/bootstrap.min.css" integrity="sha384-Gn5384xqQ1aoWXA+058RXPxPg6fy4IWvTNh0E263XmFcJlSAwiGgFAW/dAiS6JXm" crossorigin="anonymous"> <script src="https://code.jquery.com/jquery-3.2.1.slim.min.js" integrity="sha384-KJ3o2DKtIkvYIK3UENzmM7KCkRr/rE9/Qpg6aAZGJwFDMVNA/GpGFF93hXpG5KkN" crossorigin="anonymous"></script> <script src="https://cdnjs.cloudflare.com/ajax/libs/popper.js/1.12.9/umd/popper.min.js" integrity="sha384-ApNbgh9B+Y1QKtv3Rn7W3mgPxhU9K/ScQsAP7hUibX39j7fakFPskvXusvfa0b4Q" crossorigin="anonymous"></script> <script src="https://maxcdn.bootstrapcdn.com/bootstrap/4.0.0/js/bootstrap.min.js" integrity="sha384-JZR6Spejh4U02d8jOt6vLEHfe/JQGiRRSQQxSfFWpi1MquVdAyjUar5+76PVCmYl" crossorigin="anonymous"></script> <link rel="apple-touch-icon" sizes="180x180" href="images/apple-touch-icon.png"> <link rel="icon" type="image/png" sizes="32x32" href="imagesfavicon-32x32.png"> <link rel="icon" type="image/png" sizes="16x16" href="imagesfavicon-16x16.png"> <link rel="manifest" href="tools/site.webmanifest"> <link rel="icon" type="image/x-icon" href="images/favicon.ico"> <title>Maarten Sap - Home</title> </head> <script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script> <body> <nav class="navbar navbar-expand-sm navbar-dark bg-dark"> <div class="container" style="max-width: 80%;"> <a class="navbar-brand" href="index.html">Maarten Sap</a> <button class="navbar-toggler" type="button" data-toggle="collapse" data-target="#navbarNav" aria-controls="navbarNav" aria-expanded="false" aria-label="Toggle navigation"> <span class="navbar-toggler-icon"></span> </button> <div class="collapse navbar-collapse" id="navbarNav"> <ul class="navbar-nav"> <li class="nav-item"> <a class="nav-link" href="publications.html">Publications</a> </li> <li class="nav-item"> <a class="nav-link" href="contact.html">Contact</a> </li> <li class="nav-item"> <a class="nav-link" href="aboutme.html">About Me</a> </li> <li class="nav-item"> <a class="nav-link" href="cv.html">CV</a> </li> <!--li class="nav-item dropdown"> <a class="nav-link dropdown-toggle" href="" id="CVdropdown" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false"> CV </a> <div class="dropdown-menu" aria-labelledby="CVdropdown"> <a class="dropdown-item" href="cv.html" target="_blank"> html</a> <a class="dropdown-item" href="files/CVMaartenSap.pdf" target="_blank"> pdf <!--img src="images/pdf_icon.png" style="height: 1em;"-></a> </div> </li--> <!--li class="nav-item"> <a class="nav-link" href="files/CVMaartenSap.pdf" target="_blank"> <img src="images/pdf_icon_white.png" style="height: 1em;"> CV (pdf)</a> </li> <li class="nav-item"> <a class="nav-link" href="cv.html" target="_blank"> CV (html)</a> </li--> <li class="nav-item dropdown"> <a class="nav-link dropdown-toggle" href="" id="navbarDropdownMenuLink" data-toggle="dropdown" aria-haspopup="true" aria-expanded="false"> Notes/Blogposts </a> <div class="dropdown-menu" aria-labelledby="navbarDropdownMenuLink"> <a class="dropdown-item" href="notes/apply-grad-school.html">Applying to grad school</a> <a class="dropdown-item" href="notes/giving-talk-feedback.html">Giving feedback for talks</a> <a class="dropdown-item" href="notes/job-search-2020.html">Notes from my 2020 Academic Job Search</a> <a class="dropdown-item" href="notes/NLGDecoding.pptx">NLGDecoding [pptx]</a> <a class="dropdown-item" href="notes/presenting-your-research.html">Presenting your research</a> <a class="dropdown-item" href="notes/rebuttals.html">Writing rebuttals</a> <a class="dropdown-item" href="notes/turking-tips.html">Turking tips</a> <a class="dropdown-item" href="notes/writing-rec-letters.html">Writing recommendation letters</a> </div> <!--div class="dropdown-menu" aria-labelledby="navbarDropdownMenuLink"> <a class="dropdown-item" href="notes/presenting-your-research.html">Presenting Your Research</a> <a class="dropdown-item" href="notes/apply-grad-school.html">How to Apply to Grad School</a> <a class="dropdown-item" href="notes/seq2seq-tricks.html">Seq2Seq Magic Tips</a> <a class="dropdown-item" href="notes/NLGDecoding.pptx">Tutorial on beam search and NLG decoding [pptx]</a> <a class="dropdown-item" href="notes/turking-tips.html">Tips for Turking</a> <a class="dropdown-item" href="notes/rebuttals.html">Writing rebuttals</a> <a class="dropdown-item" href="notes/job-search-2020.html">Notes from my 2020 Academic Job Search</a> </div--> </li> </ul> </div> </div> </nav> <!-- Page Content --> <!--div class="clearfix hidden-sm-up"></div--> <div class="container" style="padding-top: 10px;max-width: 80%;"> <div class="row justify-content-center"> <div class="col-sm-12 col-md-12 col-lg-10"> <!-- Jumbotron Header --> <!--div class="jumbotron" style="display: inline-block"> <img src="files/M6.jpg" alt="Maarten Sap" style="padding-left:10px; margin-right: 15px; max-width:135px;float:left"> <p id="intro-text">Hi! I'm Maarten Sap, currently a PhD student at the University of Washington. I'm interested in natural language processing, specifically contextual language modelling and social science NLP applications. </p> </div--> <div class="jumbotron" style="padding: 2rem 1rem;"> <div class="container row justify-content-center"> <div class="col-lg-3 col-md-3 col-sm-12"> <img class="profile" src="images/headshots/armsCrossed_square_sm.jpg" alt="Maarten Sap" style="width: 100%; margin-bottom: 2em;"> <a class="btn btn-light" href="http://twitter.com/maartensap" target="_blank"> <img src="images/twitter-icon.png" class="icon" title="Twitter Profile" alt="Twitter Profile"> </a> <a class="btn btn-light" href="https://scholar.google.com/citations?user=gFN4QUYAAAAJ&hl=en" target="_blank"> <img src="images/google-scholar-icon.png" class="icon" title="Google Scholar Profile" alt="Google Scholar Profile"> </a> <a class="btn btn-light" href="https://www.semanticscholar.org/author/Maarten-Sap/2729164" target="_blank"> <img src="images/s2-logo-bw.png" class="icon" title="Semantic Scholar Profile" alt="Semantic Scholar Profile"> </a> </div> <div class="col-lg-9 col-md-9 col-sm-12 lead" style="padding-right:0px;"> <h2 class="display-4">Maarten Sap</h2> <p> I am an assistant professor at <a href="https://www.lti.cs.cmu.edu/" target="_blank">CMU's LTI</a> department with a courtesy appointment in <a href="https://hcii.cmu.edu/">HCII</a>, and a part-time research scientist at the <a href="https://allenai.org" target="_blank">Allen Institute for AI (AI2)</a>. My research focuses on endowing NLP systems with <strong>social intelligence and social commonsense</strong>, and understanding <strong>social inequality and bias</strong> in language. </p> <p> Before this, I was a Postdoc/Young Investigator at the <a href="https://allenai.org" target="_blank">Allen Institute for AI (AI2)</a>, working on project <a target="_blank" href="https://allenai.org/commonsense/">Mosaic</a>. I received my PhD from the University of Washington where I was advised by <a target="_blank" href="https://homes.cs.washington.edu/~nasmith/">Noah Smith</a> and <a target="_blank" href="https://homes.cs.washington.edu/~yejin/">Yejin Choi</a>, and have interned at AI2 working on social commonsense reasoning, and at Microsoft Research working on deep learning models for understanding human cognition. <br><a href="bio.html" style="font-size: .95rem;">[bio for talks]</a> </p> <!--p </p--> </div> <div class="col-12" style="background: white; padding: 4px; border-radius: 5px 5px; font-size: 1.15rem; font-weight: 300;"> <h5><em>Recent updates:</em></h5> <p style="margin-bottom: .5em;"><em><strong>November 2024</em></strong> 🫂👨‍🏫: Very excited that I now have a courtesy appointment in the <a target="_blank" href="https://hcii.cmu.edu/">Human Computer Interaction Institute</a>!</p> <p style="margin-bottom: .5em;"><em><strong>November 2024</em></strong> 🔍🧑‍🎓: As a reminder, due to my lab being quite full already, I'm not taking any students in this upcoming PhD application cycle 😟.</p> <p style="margin-bottom: .5em;"><em><strong>November 2024</em></strong> 🏖️📚: Excited to give a talk at the <a target="_blank" href="https://sites.google.com/cs.stonybrook.edu/wnu2024">6th Workshop on Narrative Understanding</a> on <em>Computational Methods of Social Causes and Effects of Stories</em>.</p> <p style="margin-bottom: .5em;"><em><strong>November 2024</em></strong> 🏖️🐊: Excited to attend <a target="_blank" href="https://2024.emnlp.org/">EMNLP</a> in Miami, where my students will be presenting their papers: <a target="_blank" href="https://joelmire.notion.site/">Joel Mire</a> on <a target="_blank" href="https://maartensap.com/publications.html#mire2024empirical"><em>The Empirical Variability of Narrative Perceptions of Social Media Texts</em></a>, <a target="_blank" href="https://jocelynshen.com/">Jocelyn Shen</a> on <a target="_blank" href="https://arxiv.org/abs/2405.17633"><em>HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs</em></a>, and <a target="_blank" href="https://xuhuizhou.github.io/">Xuhui Zhou</a> on <a target="_blank" href="http://arxiv.org/abs/2403.05020"><em>Is This the Real Life? Is This Just Fantasy? The Misleading Success of Simulating Social Interactions With LLMs</em></a>.</p> <p style="margin-bottom: .5em;"><em><strong>November 2024</em></strong> 🧑‍🎓👨🏼‍🏫: Giving a talk at the University of Pittsburgh CS colloquium on <em>Artificial Social Intelligence? On the challenges of Socially Aware and Ethically informed LLMs</em> (Fall 2024 version). Recording is on <a target="_blank" href="https://youtu.be/wiGctXJAKcc">Youtube</a>.</p> <p style="margin-bottom: .5em;"><em><strong>October 2024</em></strong> 🗽🦁: Giving a talk in the Columbia NLP seminar on <em>Artificial Social Intelligence? On the challenges of Socially Aware and Ethically informed LLMs</em> (Fall 2024 version).</p> <p style="margin-bottom: .5em;"><em><strong>October 2024</em></strong> 🔔: Excited to attend <a target="_blank" href="https://colmweb.org/">COLM</a> at UPenn in Philadelphia, and re-visiting my old haunts around UPenn (from 10 years ago).</p> <!--p style="margin-bottom: .5em;"> <em><strong>December 2021 update</strong></em>🧑‍🎓:I will likely be taking students this coming PhD application cycle. If you're interested in working with me on <em>social commonsense, social biases in language, or ethics in AI</em>, please apply to <a href="https://www.lti.cs.cmu.edu/apply-lti" target="_blank">CMU's LTI</a>. </p> <p style="margin-bottom: .5em;"> <em><strong>July 2021 update</strong></em>👨🏼‍🎓: I successfully defended my PhD thesis titled <em>Positive AI with Social Commonsense Models</em> (<a target="_blank" href="pdfs/sap2021positiveAIwithSocialCommonsenseModels.pdf">read the thesis here</a>, or <a href="https://www.youtube.com/watch?v=VKd4Y7Ykh6o" target="_blank">watch the recording here<a/>). Thanks to my advisors, committee, and everyone who attended! </p> <p style="margin-bottom: .5em;"> <em><strong>May 2021 update</strong></em>🥳: I will be joining <a href="https://www.lti.cs.cmu.edu/" target="_blank">CMU's LTI</a> department as an assistant professor👨🏼‍🏫in Fall 2022. If you wish to work with me, <a href="contact.html">see the "contact" page</a>. Before starting there, I will be a postdoc at <a href="https://allenai.org" target="_blank">AI2</a> on project <a target="_blank" href="https://allenai.org/commonsense/">Mosaic</a> 👨🏼‍🔬 starting Fall 2021. </p--> <p><small style="font-weight: 300;"><a target="_blank" href="news.html">[older news]</a></small></p> </div> </div> </div> <hr> <!-- Title --> <!--div class="row"> <div class="col-lg-12"> <h3></h3> </div> </div--> <!-- /.row --> <!-- Page Features --> <div class="row"> <div class="col-12" id="research-group"> <h2>My research group:</h2> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/dan-chechelnitsky.png"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://chechelnitskd.github.io/">Dan Chechelnitsky</a></h5> <p class="card-text font-italic"> LTI PhD student <br><small>co-advised with <a href="https://scholar.google.com/citations?user=S5NGkFsAAAAJ&hl=en&oi=ao">Chrysoula Zerva</a></small></h5> </p> </div> </div> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/joel-mire.jpg"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://joelmire.notion.site/">Joel Mire</a></h5> <p class="card-text font-italic"> LTI MLT student </p> </div> </div> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/karina-halevy.jpg"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://enscma2.github.io/">Karina Halevy</a></h5> <p class="card-text font-italic"> LTI PhD student <br><small>co-advised with <a href="https://scholar.google.com.vn/citations?user=-y6SIhQAAAAJ&hl=vi">Mona Diab</a></small> </p> </div> </div> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/jimin-mun.jpeg"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://jiminmun.github.io/">Jimin Mun</a></h5> <p class="card-text font-italic"> LTI PhD student </p> </div> </div> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/jocelyn-shen.jpg"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://jocelynshen.com/">Jocelyn Shen</a></h5> <p class="card-text font-italic"> MIT PhD student <br><small>co-advised with <a href="https://www.media.mit.edu/people/cynthiab/overview/">Cynthia Breazeal</a></small> </p> </div> </div> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/akhila-yerukola.jpg"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://akhila-yerukola.github.io/" target="_blank">Akhila Yerukola</a></h5> <p class="card-text font-italic"> LTI PhD student </p> </div> </div> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/mingqian-zheng.jpg"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://eeelisa.github.io/" target="_blank">Mingqian Zheng</a></h5> <p class="card-text font-italic"> LTI PhD student <br><small>co-advised with <a href="https://www.cs.cmu.edu/~cprose/">Carolyn Rosé</a></small> </p> </div> </div> </div> <div class="col-lg-3 col-xl-3 col-sm-4 col-md-4"> <div class="card no-border"> <img class="card-img-profile" src="images/students/xuhui-zhou.jpg"/> <div class="card-body text-center"> <h5 class="card-title"><a href="https://xuhuizhou.github.io/" target="_blank">Xuhui Zhou</a></h5> <p class="card-text font-italic"> LTI PhD student </p> </div> </div> </div> </div> <hr> <div class="row"> <div class="col-12"> <h2>Overarching Research Themes</h2> </div> <div class="col-12" id="content"> *Extracted by GPT-4, there may be inconsistencies.* #### *Ethics in AI Development* My research group explores the complex dimensions of ethical AI and its implications for society. One of our pivotal works, [HAICOSYSTEM: An Ecosystem for Sandboxing Safety Risks in Human-AI Interactions](http://arxiv.org/abs/2409.16427), introduces a framework designed to simulate potential safety risks associated with human-AI interactions. We also investigate the balance between usability and ethical considerations through our paper, [AI-LieDar: Examine the Trade-off Between Utility and Truthfulness in LLM Agents](https://arxiv.org/abs/2409.09013), which highlights the dilemmas faced by LLMs in delivering truthful responses. Our efforts culminate in [Particip-AI: A Democratic Surveying Framework](https://arxiv.org/abs/2403.14791) that aims to engage a broader audience in discussions of future AI applications and their impacts. #### *Narrative Dynamics in AI* My research group explores the way narratives are constructed and perceived through AI technologies. We delve into emotional resonance in storytelling with our important study, [HEART-felt Narratives: Tracing Empathy and Narrative Style in Personal Stories with LLMs](https://arxiv.org/abs/2405.17633), which examines how narrative styles can evoke empathy. Additionally, our work, [Modeling Empathic Similarity in Personal Narratives](https://arxiv.org/abs/2305.14246), quantifies connections between audiences and narratives, enriching our understanding of storytelling mechanics. Another significant contribution, [Quantifying the narrative flow of imagined versus autobiographical stories](https://www.pnas.org/doi/10.1073/pnas.2211715119), provides insights into how different types of personal stories affect listener engagement. #### *Social Intelligence and Simulation* My research group explores how social intelligence can be embedded and evaluated in language models. We critically assess the limitations of simulating genuine interpersonal engagements in our study, [Is This the Real Life? Is This Just Fantasy? The Misleading Success of Simulating Social Interactions With LLMs](http://arxiv.org/abs/2403.05020), which reveals the gaps between simulated responses and authentic human communication. To foster more effective models, we research the efficacy of context in understanding social dynamics through [Clever Hans or Neural Theory of Mind? Stress Testing Social Reasoning in Large Language Models](https://arxiv.org/abs/2305.14763). Our ongoing efforts also led to [SOTOPIA: Interactive Evaluation for Social Intelligence in Language Agents](https://arxiv.org/abs/2310.11667), which advances the measures of social reasoning capabilities among AI agents. #### *Addressing Toxic Language* My research group explores the pressing challenges in mitigating toxic language generation within AI systems. We highlight multilingual aspects of this issue with our notable study, [PolygloToxicityPrompts: Multilingual Evaluation of Neural Toxic Degeneration in Large Language Models](https://arxiv.org/abs/2405.09373), which scrutinizes how toxic content manifests across languages. Another critical observation is presented in [Counterspeakers’ Perspectives: Unveiling Barriers and AI Needs in the Fight against Online Hate](https://arxiv.org/abs/2403.00179), offering insights into the requirements for combating hate speech effectively. Additionally, our work, [Leftover-Lunch: Advantage-based Offline Reinforcement Learning for Language Models](https://arxiv.org/abs/2305.14718), investigates reinforcement learning strategies to enhance ethical dialogue generation. </div> <!--div class="row col-12 text-center"> <div class="col-12 col-sm-12"> <div class="card" style="border: solid #1b9e77"> <h5 class="card-title card-header" style="background: #b3e2cd">Detecting and Mitigating Social Biases in Language</h5> <div class=""> <img class="card-img-top" src="images/SBF_thumbnail.png"/> <img class="card-img-top" src="images/powerTransformer.png"> </div> <div class="card-body"> <p class="card-text"> Language can perpetuate social biases and toxicity against oppressed or marginalized groups. I want to investigate new ways of representing and detecting such harmful content in text (e.g., <a href="social-bias-frames/" target="_blank">Social Bias Frames</a>). Additionally, I want to harness NLP systems to combat stereotypical or harmful statements in language, through controllable text generation (e.g., with <a href="https://aclanthology.org/2021.acl-long.522/" target="_blank">DExperts</a>) or controllable text debiasing (e.g., with <a href="controllable-debiasing/" target="_blank">PowerTransformer</a>). </p> <p> <em>In the future</em>, I want to make this technology more context-aware and human-centric, e.g., by incorporating power differentials between speaker and listener, and studying human-in-the-loop methods for toxicity detection or text debiasing. </p> </div> </div> </div> <div class="col-12 col-sm-12"> <div class="card" style="border: solid #7570b3;"> <h5 class="card-title card-header" style="background: #cbd5e8;">Commonsense Reasoning for Socially Aware NLP</h5> <div class=""> <img class="card-img-top" src="images/ATOMIC-fullGraph.png"/> <img class="card-img-top" src="images/SocialIQa_weblogo.png"> </div> <div class="card-body"> <p class="card-text"> Through theory-of-mind, Humans are trivially able to reason about other people's intents and reactions to everyday situations. I am interested in studying how AI systems can do this type of social commonsense reasoning. For example, this requires giving models knowledge of social commensense (e.g., with <a href="https://uwnlp.github.io/event2mind/" target="_blank">Event2Mind</a> or <a href="https://homes.cs.washington.edu/~msap/atomic/" target="_blank">ATOMIC</a>, and methods like <a href="https://mosaickg.apps.allenai.org/model_comet2020_people_events" target="_blank">CoMET</a>) or social acceptibility (<a href="https://maxwellforbes.com/social-chemistry/" target="_blank">Social Chemistry</a>). Additionally, this requires creating benchmarks for measuring models' social commonsense abilities (e.g., with <a href="https://leaderboard.allenai.org/socialiqa/submissions/get-started" target="_blank">Social IQa</a>, or <a href="https://uwnlp.github.io/storycommonsense/" target="_blank">Story Commonsense</a>). </p> <p> <em>In the future</em>, I want to keep investigating this elusive goal of machine social commonsense, which <a href="publications.html#sap2022neuralToM">we show is still not present in LLMs like GPT-3</a>. Additionally, I want to explore positive applications of this research, e.g., for therapeutic setting or for helping people with cognitive disabilities. </p> </div> </div> </div> <div class="col-12 col-sm-12"> <div class="card" style="border: solid #d95f02;"> <h5 class="card-title card-header" style="background: #fdcdac">Analyzing the Ethics and Transparency of AI models</h5> <div class=""> <img class="card-img-top" src="images/twitterAAE.png"/> <img class="card-img-top" src="images/c4-fig1-v2.png"> </div> <div class="card-body"> <p class="card-text"> AI and NLP systems unfortunately encode social biases and stereotypes. I'm passionate about analyzing and diagnosing the potential negative societal impacts of these systems. For example, I've uncovered <a href="publications.html#sap2019risk" target="_blank">severe racial bias in hate speech detection datasets and models</a>, and subsequently analyzed <a href="publications.html#zhou2021challenges" target="_blank">whether robustness methods for NLP can mitigate</a> them, as well as understanding the <a href="publications.html#sap2022annotatorsWithAttitudes" target="_blank">psychological attitudes that cause over- and under-detection of content as toxic</a>. Additionally, I've scrutinized recent pretrained language models and their training data with respect to biases, toxicity, and fake news (e.g., <a href="https://toxicdegeneration.allenai.org/" target="_blank">measuring GPT-2 and GPT-3's neural toxic degeneration</a>, and <a href="publications.html#dodge2021documentingC4" target="_blank">documenting the English C4 Webtext Crawl</a>). </p> <p> <em>In the future</em>, I plan to keep diagnosing and mitigating the ethical, fairness, and representation issues in AI systems, especially from a human-centric perspective of end-users and other stakeholders. </p> </div> </div> </div> <div class="col-12 col-sm-12"> <div class="card" style="border: solid #984ea3;"> <h5 class="card-title card-header" style="background: #decbe4">Socially Aware Conversational AI</h5> <div class=""> <img class="card-img-top" src="images/toxichat.png"> <img class="card-img-top" src="images/prosocialDialog.png"> </div> <div class="card-body"> <p class="card-text"> Conversational agents are one of the most important ways that humans will interface with AI systems, whether they are talk with AI systems (e.g., chatbots) or chatting with each others with AI-assisted interfaces (e.g., smart suggestions). I'm interested in studying how we can make conversational agents more socially aware and better respond to toxicity, based on my experience winning the <a href="https://sounding-board.github.io/">inaugural Alexa Prize</a> in 2017. For example, in our <a href="publications.html#baheti2021justSayNo" target="_blank">ToxiChat paper</a>, we quantified how chatbots often agree with toxic utterances more than with neutral comments. I've also developed large-scale datasets for developing dialogue agents that are more prosocial (<a href="publications.html#kim2022prosocialDialog">ProSocial Dialogs</a>) and socially competent (with <a href="publications.html#kim2022soda">SODA</a>). </p> <p> <em>In the future</em>, I am interested in understanding how conversational agents can better reason about communicative intent when responding, how communication can be learned in an emergent way, and how we can use reinforcement learning to make dialog agents better. </p> </div> </div> </div> </div--> </div> <!-- /.row --> <div class="row" style="margin-top: 20px;"> <div class="col-12"><h3>More about me</h3></div> <div class="col-12 btn-toolbar"role="toolbar"> <div class="btn-group mr-2"> <a href="publications.html" class="btn btn-light">Browse my publications</a> </div> <div class="btn-group mr-2"> <a href="contact.html" class="btn btn-light">Contact me</a> </div> <div class="btn-group mr-2"> <a href="aboutme.html" class="btn btn-light">More about me</a> </div> <div class="btn-group mr-2"> <a href="cv.html" class="btn btn-light" target="_blank"> CV</a> </div> </div> </div> </div> <!--div class="clearfix hidden-sm-up"></div--> <!--div class="col-sm-12 col-md-12 col-lg-3"> <div class="row justify-content-center" style="margin-top: 10px;"> <!--div class="col-sm-8 col-lg-12 twitter-timeline-maarten"> <a class="twitter-timeline" href="https://twitter.com/MaartenSap" data-widget-id="659897450853830656" style="display:inline-block">Tweets by @MaartenSap</a> <script> ! function(d, s, id) { var js, fjs = d.getElementsByTagName(s)[0], p = /^http:/.test(d.location) ? 'http' : 'https'; if (!d.getElementById(id)) { js = d.createElement(s); js.id = id; js.src = p + "://platform.twitter.com/widgets.js"; fjs.parentNode.insertBefore(js, fjs); } }(document, "script", "twitter-wjs"); </script> </div--> <!--div class="col-sm-4 col-lg-12"> <h4>More about me</h4> <a href="publications.html" class="btn btn-margin btn-secondary">Publications</a> <a href="contact.html" class="btn btn-margin btn-secondary">Contact me</a> <a href="aboutme.html" class="btn btn-margin btn-secondary">More about me</a> <a href="/~msap/files/CVMaartenSap.pdf" class="btn btn-margin btn-secondary" target="_blank">CV</a> <!--div class="btn-toolbar" role="toolbar"> <div class="btn-group mr-2"> <a href="publications.html" class="btn btn-light">Browse through publications</a> </div> <div class="btn-group mr-2"> <a href="contact.html" class="btn btn-light">Contact me</a> </div> <div class="btn-group mr-2"> <a href="aboutme.html" class="btn btn-light">More about me</a> </div> </div--> <!--/div> </div> </div--> </div> </div> </body> <!-- Footer --> <footer class="footer"> <hr> <div style="width:99%"> <!--div class="wraptocenter"--> <!--div class="col-sm-12"> <img src="files/AllenSchool.png" style="width:200px;"> </div--> <div class="row"> <div class="col-2"> </div> <div class="col-8 text-center"> <img src="images/LTI-logo.png" style="width:300px; margin-bottom: 20px;"> </div> <div class="col-2 mb-1"> <div class="text-muted small" style="font-style: italic; text-align: right;"> Last updated: <br> 2024-11-22 </div> </div> </div> </div> </footer> </html> <script> document.getElementById('content').innerHTML = marked.parse(document.getElementById('content').innerHTML); </script> <style> .card { padding: 0px; margin: 10px 5px; } .card-img-profile { margin: auto; border-radius: .25em; max-height: 300px; max-width: 300px; height: 100%; width: 100%; } .card-img-top { max-height: 170px; max-width: 70%; width: auto; margin-left: 1em; margin-right: 1em; } .btn { margin-top: .2em; } .profile { border-radius: .1em; box-shadow: 0 4px 8px 0 rgba(0, 0, 0, 0.2), 0 6px 20px 0 rgba(0, 0, 0, 0.19); } .no-border{ border: none; } </style>

Pages: 1 2 3 4 5 6 7 8 9 10