More Like this

1. Integrating Dynamic Supports into an Equity Teaching Simulation to Promote Equity Mindsets

   https://doi.org/10.1145/3491140.3528327  

   Marvez, G. R.; Zheng, Tianyuan; Littenberg-Tobias, Joshua; Hillaire, Garron; O'Brien, Sara; Reich, Justin (June 2022, L@S '22: Proceedings of the Ninth ACM Conference on Learning @ Scale)

   Implementing high-quality professional learning on diversity, equity, and inclusion (DEI) issues is a massive scaling challenge. Integrating dynamic support using natural language processing (NLP) into equity teaching simulations may allow for more responsive, personalized training in this field. In this study, we trained machine learning models on participants’ text responses in an equity teaching simulation (494 users; 988 responses) to detect certain text features related to equity. We then integrated these models into the simulation to provide dynamic supports to users during the simulation. In a pilot study (N = 13), we found users largely thought the feedback was accurate and incorporated the feedback in subsequent simulation responses. Future work will explore replicating these results with larger and more representative samples 

   more » « less
2. Persuasion of the Undecided: Language vs. the Listener.

   Longpre, Liane; Durmus, Esin; and Cardie, Claire (January 2019, Proceedings of the 6th Workshop on Argument Mining)

   This paper examines the factors that govern persuasion for a priori UNDECIDED versus DECIDED audience members in the context of on-line debates. We separately study two types of influences: linguistic factors — features of the language of the debate itself; and audience factors — features of an audience member encoding demographic information, prior beliefs, and debate platform behavior. In a study of users of a popular debate platform, we find first that different combinations of linguistic features are critical for predicting persuasion outcomes for UNDECIDED versus DECIDED members of the audience. We additionally find that audience factors have more influence on predicting the side (PRO/CON) that persuaded UNDECIDED users than for DECIDED users that flip their stance to the opposing side. Our results emphasize the importance of considering the undecided and decided audiences separately when studying linguistic factors of persuasion. 

   more » « less
3. “Any Other Thoughts, Hedgehog?” Linking Deliberation Chains in Collaborative Dialogues

   https://doi.org/10.18653/v1/2024.findings-emnlp.305  

   Nath, Abhijnan; Venkatesha, Videep; Bradford, Mariah; Chelle, Avyakta; Youngren, Austin C; Mabrey, Carlos; Blanchard, Nathaniel; Krishnaswamy, Nikhil (January 2024, Association for Computational Linguistics)

   Question-asking in collaborative dialogue has long been established as key to knowledge construction, both in internal and collaborative problem solving. In this work, we examine probing questions in collaborative dialogues: questions that explicitly elicit responses from the speaker`s interlocutors. Specifically, we focus on modeling the causal relations that lead directly from utterances earlier in the dialogue to the emergence of the probing question. We model these relations using a novel graph-based framework of *deliberation chains*, and realize the problem of constructing such chains as a coreference-style clustering problem. Our framework jointly models probing and causal utterances and the links between them, and we evaluate on two challenging collaborative task datasets: the Weights Task and DeliData. Our results demonstrate the effectiveness of our theoretically-grounded approach compared to both baselines and stronger coreference approaches, and establish a standard of performance in this novel task. 

   more » « less
4. Diverse and faithful knowledge-grounded dialogue generation via sequential posterior inference

   Xu, Y.; Kong, D.; Xu, D.; Ji, Z.; Pang, B.; Fung, P.; Wu, Y. N. (October 2023, International Conference on Machine Learning (ICML 2023))

   The capability to generate responses with diversity and faithfulness using factual knowledge is paramount for creating a human-like, trustworthy dialogue system. Common strategies either adopt a two-step paradigm, which optimizes knowledge selection and response generation separately and may overlook the inherent correlation between these two tasks, or leverage conditional variational method to jointly optimize knowledge selection and response generation by employing an inference network. In this paper, we present an end-to-end learning framework, termed Sequential Posterior Inference (SPI), capable of se- lecting knowledge and generating dialogues by approximately sampling from the posterior distribution. Unlike other methods, SPI does not require the inference network or assume a simple geometry of the posterior distribution. This straightforward and intuitive inference procedure of SPI directly queries the response generation model, allowing for accurate knowledge selection and generation of faithful responses. In addition to modeling contributions, our experimental results on two common dialogue datasets (Wizard of Wikipedia and Holl-E) demonstrate that SPI outperforms previous strong baselines according to both automatic and human evaluation metrics. 

   more » « less
5. “Nice Try, Kiddo”: Investigating Ad Hominems in Dialogue Responses

   https://doi.org/10.18653/v1/2021.naacl-main.60  

   Sheng, Emily; Chang, Kai-Wei; Natarajan, Prem; Peng, Nanyun (January 2021, Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies)

   null (Ed.)

   Ad hominem attacks are those that target some feature of a person’s character instead of the position the person is maintaining. These attacks are harmful because they propagate implicit biases and diminish a person’s credibility. Since dialogue systems respond directly to user input, it is important to study ad hominems in dialogue responses. To this end, we propose categories of ad hominems, compose an annotated dataset, and build a classifier to analyze human and dialogue system responses to English Twitter posts. We specifically compare responses to Twitter topics about marginalized communities (#BlackLivesMatter, #MeToo) versus other topics (#Vegan, #WFH), because the abusive language of ad hominems could further amplify the skew of power away from marginalized populations. Furthermore, we propose a constrained decoding technique that uses salient n-gram similarity as a soft constraint for top-k sampling to reduce the amount of ad hominems generated. Our results indicate that 1) responses from both humans and DialoGPT contain more ad hominems for discussions around marginalized communities, 2) different quantities of ad hominems in the training data can influence the likelihood of generating ad hominems, and 3) we can use constrained decoding techniques to reduce ad hominems in generated dialogue responses. 

   more » « less