Saloni Dash

Abbavaram Gowtham Reddy, Saloni Dash, Amit Sharma et al.

A machine learning model, under the influence of observed or unobserved confounders in the training data, can learn spurious correlations and fail to generalize when deployed. For image classifiers, augmenting a training dataset using counterfactual examples has been empirically shown to break spurious correlations. However, the counterfactual generation task itself becomes more difficult as the level of confounding increases. Existing methods for counterfactual generation under confounding consider a fixed set of interventions (e.g., texture, rotation) and are not flexible enough to capture diverse data-generating processes. Given a causal generative process, we formally characterize the adverse effects of confounding on any downstream tasks and show that the correlation between generative factors (attributes) can be used to quantitatively measure confounding between generative factors. To minimize such correlation, we propose a counterfactual generation method that learns to modify the value of any attribute in an image and generate new images given a set of observed attributes, even when the dataset is highly confounded. These counterfactual images are then used to regularize the downstream classifier such that the learned representations are the same across various generative factors conditioned on the class label. Our method is computationally efficient, simple to implement, and works well for any number of generative factors and confounding variables. Our experimental results on both synthetic (MNIST variants) and real-world (CelebA) datasets show the usefulness of our approach.

Saloni Dash, Amélie Reymond, Emma S. Spiro et al. · uw

Reasoning in humans is prone to biases due to underlying motivations like identity protection, that undermine rational decision-making and judgment. This motivated reasoning at a collective level can be detrimental to society when debating critical issues such as human-driven climate change or vaccine safety, and can further aggravate political polarization. Prior studies have reported that large language models (LLMs) are also susceptible to human-like cognitive biases, however, the extent to which LLMs selectively reason toward identity-congruent conclusions remains largely unexplored. Here, we investigate whether assigning 8 personas across 4 political and socio-demographic attributes induces motivated reasoning in LLMs. Testing 8 LLMs (open source and proprietary) across two reasoning tasks from human-subject studies -- veracity discernment of misinformation headlines and evaluation of numeric scientific evidence -- we find that persona-assigned LLMs have up to 9% reduced veracity discernment relative to models without personas. Political personas specifically, are up to 90% more likely to correctly evaluate scientific evidence on gun control when the ground truth is congruent with their induced political identity. Prompt-based debiasing methods are largely ineffective at mitigating these effects. Taken together, our empirical findings are the first to suggest that persona-assigned LLMs exhibit human-like motivated reasoning that is hard to mitigate through conventional debiasing prompts -- raising concerns of exacerbating identity-congruent reasoning in both LLMs and humans.

Alejandro Cuevas, Saloni Dash, Bharat Kumar Nayak et al.

Disinformation is among the top risks of generative artificial intelligence (AI) misuse. Global adoption of generative AI necessitates red-teaming evaluations (i.e., systematic adversarial probing) that are robust across diverse languages and cultures, but red-teaming datasets are commonly US- and English-centric. To address this gap, we propose "anecdoctoring", a novel red-teaming approach that automatically generates adversarial prompts across languages and cultures. We collect misinformation claims from fact-checking websites in three languages (English, Spanish, and Hindi) and two geographies (US and India). We then cluster individual claims into broader narratives and characterize the resulting clusters with knowledge graphs, with which we augment an attacker LLM. Our method produces higher attack success rates and offers interpretability benefits relative to few-shot prompting. Results underscore the need for disinformation mitigations that scale globally and are grounded in real-world adversarial misuse.

Yiwei Xu, Saloni Dash, Sungha Kang et al.

This study examined how AI-generated summaries, which have become visually prominent in online search results, affect how users think about different issues. In a preregistered randomized controlled experiment, participants (N = 2,004) viewed mock search result pages varying in the presence (vs. absence), placement (top vs. middle), and stance (benefit-framed vs. harm-framed) of AI-generated summaries across four publicly debated topics. Compared to a no-summary control group, participants exposed to AI-generated summaries reported issue attitudes, behavioral intentions, and policy support that aligned more closely with the AI summary stance. The summaries placed at the top of the page produced stronger shifts in users' issue attitudes (but not behavioral intentions or policy support) than those placed at the middle of the page. We also observed moderating effects from issue familiarity and general trust toward AI. In addition, users perceived the AI summaries more useful when it emphasized health harms versus benefits. These findings suggest that AI-generated search summaries can significantly shape public perceptions, raising important implications for the design and regulation of AI-integrated information ecosystems.

Abbavaram Gowtham Reddy, Saketh Bachu, Saloni Dash et al.

Counterfactual data augmentation has recently emerged as a method to mitigate confounding biases in the training data. These biases, such as spurious correlations, arise due to various observed and unobserved confounding variables in the data generation process. In this paper, we formally analyze how confounding biases impact downstream classifiers and present a causal viewpoint to the solutions based on counterfactual data augmentation. We explore how removing confounding biases serves as a means to learn invariant features, ultimately aiding in generalization beyond the observed data distribution. Additionally, we present a straightforward yet powerful algorithm for generating counterfactual images, which effectively mitigates the influence of confounding effects on downstream classifiers. Through experiments on MNIST variants and the CelebA datasets, we demonstrate how our simple augmentation method helps existing state-of-the-art methods achieve good results.

Saloni Dash, Dibyendu Mishra, Gazal Shekhawat et al.

Influencers are key to the nature and networks of information propagation on social media. Influencers are particularly important in political discourse through their engagement with issues, and may derive their legitimacy either solely or in large part through online operation, or have an offline sphere of expertise such as entertainers, journalists etc. To quantify influencers' political engagement and polarity, we use Google's Universal Sentence Encoder (USE) to encode the tweets of 6k influencers and 26k Indian politicians during political crises in India. We then obtain aggregate vector representations of the influencers based on their tweet embeddings, which alongside retweet graphs help compute their stance and polarity with respect to these political issues. We find that influencers engage with the topics in a partisan manner, with polarized influencers being rewarded with increased retweeting and following. Moreover, we observe that specific groups of influencers are consistently polarized across all events. We conclude by discussing how our study provides insights into the political schisms of present-day India, but also offers a means to study the role of influencers in exacerbating political polarization in other contexts.

Saloni Dash, Vineeth N Balasubramanian, Amit Sharma

Counterfactual examples for an input -- perturbations that change specific features but not others -- have been shown to be useful for evaluating bias of machine learning models, e.g., against specific demographic groups. However, generating counterfactual examples for images is non-trivial due to the underlying causal structure on the various features of an image. To be meaningful, generated perturbations need to satisfy constraints implied by the causal model. We present a method for generating counterfactuals by incorporating a structural causal model (SCM) in an improved variant of Adversarially Learned Inference (ALI), that generates counterfactuals in accordance with the causal relationships between attributes of an image. Based on the generated counterfactuals, we show how to explain a pre-trained machine learning classifier, evaluate its bias, and mitigate the bias using a counterfactual regularizer. On the Morpho-MNIST dataset, our method generates counterfactuals comparable in quality to prior work on SCM-based counterfactuals (DeepSCM), while on the more complex CelebA dataset our method outperforms DeepSCM in generating high-quality valid counterfactuals. Moreover, generated counterfactuals are indistinguishable from reconstructed images in a human evaluation experiment and we subsequently use them to evaluate the fairness of a standard classifier trained on CelebA data. We show that the classifier is biased w.r.t. skin and hair color, and how counterfactual regularization can remove those biases.

Saloni Dash, Ritik Dutta, Isabelle Guyon et al.

Synthetic medical data which preserves privacy while maintaining utility can be used as an alternative to real medical data, which has privacy costs and resource constraints associated with it. At present, most models focus on generating cross-sectional health data which is not necessarily representative of real data. In reality, medical data is longitudinal in nature, with a single patient having multiple health events, non-uniformly distributed throughout their lifetime. These events are influenced by patient covariates such as comorbidities, age group, gender etc. as well as external temporal effects (e.g. flu season). While there exist seminal methods to model time series data, it becomes increasingly challenging to extend these methods to medical event time series data. Due to the complexity of the real data, in which each patient visit is an event, we transform the data by using summary statistics to characterize the events for a fixed set of time intervals, to facilitate analysis and interpretability. We then train a generative adversarial network to generate synthetic data. We demonstrate this approach by generating human sleep patterns, from a publicly available dataset. We empirically evaluate the generated data and show close univariate resemblance between synthetic and real data. However, we also demonstrate how stratification by covariates is required to gain a deeper understanding of synthetic data quality.