Estimating Causal Effects of Tone in Online Debates
Authors: Dhanya Sridhar, Lise Getoor
IJCAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper, we estimate the causal effect of reply tones in debates on linguistic and sentiment changes in subsequent responses. We study debates from 4FORUMS.COM and compare annotated tones of replying such as emotional versus factual, or reasonable versus attacking. We show that our latent confounder representation reduces bias in ATE estimation. Our results suggest that factual and asserting tones affect dialogue and provide a methodology for estimating causal effects from text. |
| Researcher Affiliation | Academia | Dhanya Sridhar1 and Lise Getoor2 1Columbia University 2UC Santa Cruz ds3778@columbia.edu |
| Pseudocode | No | The paper describes models and processes but does not include structured pseudocode or algorithm blocks. |
| Open Source Code | Yes | Code and data to reproduce all results are available. 1 github.com/dsridhar91/debate-causal-effects |
| Open Datasets | Yes | To estimate the ATE of tone, we use the 4FORUMS.COM corpus collected and annotated as part of the Internet Argument Corpus [Walker et al., 2012b]. |
| Dataset Splits | Yes | We perform five-fold crossvalidation on the triples. |
| Hardware Specification | No | The paper does not specify any hardware details used for running experiments. |
| Software Dependencies | No | The paper mentions tools like LIWC and LDA but does not provide specific version numbers for any software dependencies. |
| Experiment Setup | Yes | We fit LDA with k = 50 topics. For each topic, this yields a document-term-frequency matrix of roughly 30, 000 posts and 400 remaining terms after preprocessing. We fit the propensity score P(T = 1|Z) (used by ˆψIPW, ˆψAIPW) with logistic regression using the observed treatments and constructed confounder representations. We fit the expected outcomes, Q(Z, T = 0), Q(Z, T = 1) (used by ˆψMLE, ˆψAIPW) with linear regression. |