Demonstration of 'ARC:AI Rating through Causality' tool
Collaborators: AIISC, Department of Computer Science and Engineering - University of South Carolina
Contributors: Kausik Lakkaraju, Likitha Valluru, Biplav Srivastava, Marco Valtorta
We introduce ARC, a tool to rate AI systems for bias through a causal lens.
The main objective of the tool is to assist developers in building better models
and aid end-users in making informed decisions based on the available data.
The tool is extensible and currently supports three different AI tasks: binary classification,
sentiment analysis, and group recommendation. It gives users the option of choosing data
for a task and rating AI systems for bias with respect to different protected attributes
present in the data. The rating method is system-independent and the ratings given by
the algorithm are causally interpretable. These ratings help the user make an informed
decision based on the data in hand. The demonstration video is available here:
Rating Sentiment Analysis Systems for Bias through a Causal Lens
Collaborators: AIISC, Department of Computer Science and Engineering - University of South Carolina
Contributors: Kausik Lakkaraju, Biplav Srivastava, Marco Valtorta
Sentiment Analysis Systems (SASs) are data-driven Artificial Intelligence (AI) systems that assign one or more numbers to convey the polarity and emotional intensity of a given piece of text. However, like other automatic machine learning systems, SASs can exhibit model uncertainty, resulting in drastic swings in output with even small changes in input. This issue becomes more problematic when inputs involve protected attributes like gender or race, as it can be perceived as bias or unfairness. To address this, we propose a novel method to assess and rate SASs. We perturb inputs in a controlled causal setting to test if the output sentiment is sensitive to protected attributes while keeping other components of the textual input, such as chosen emotion words, fixed. Based on the results, we assign labels (ratings) at both fine-grained and overall levels to indicate the robustness of the SAS to input changes. The ratings can help decision-makers improve online content by reducing hate speech, often fueled by biases related to protected attributes such as gender and race. These ratings provide a principled basis for comparing SASs and making informed choices based on their behavior. The ratings also benefit all users, especially developers who reuse off-the-shelf SASs to build larger AI systems but do not have access to their code or training data to compare.
Representative Publications
- K. Lakkaraju, B. Srivastava and M. Valtorta, "Rating Sentiment Analysis Systems for Bias Through a Causal Lens," in IEEE Transactions on Technology and Society, doi: 10.1109/TTS.2024.3375519.
keywords: {AI;Bias;Causality;Rating;Sentiment Analysis Systems;User Trust}.
[Paper] [GitHub Respository] [BibTex]
The Effect of Human v/s Synthetic Test Data and Round-tripping on Assessment of Sentiment Analysis Systems for Bias
Collaborators: AIISC, Department of Computer Science and Engineering, Department of Integrated Inforamtion Technology - University of South Carolina, Netaji Subhas University of Technology
Contributors: Kausik Lakkaraju, Aniket Gupta, Biplav Srivastava, Marco Valtorta, Dezhi Wu
Sentiment Analysis Systems (SASs) are data-driven Artificial Intelligence (AI) systems that output polarity and emotional intensity when given a piece of text as input. Like other AIs, SASs are also known to have unstable behavior when subjected to changes in data which can make them problematic to trust out of concerns like bias when AI works with humans and data has protected attributes like gender, race, and age. Recently, an approach was introduced to assess SASs in a blackbox setting without training data or code, and rating them for bias using synthetic English data. We augment it by introducing two human-generated chatbot datasets and also considering a round-trip setting of translating the data from one language to the same through an intermediate language. We find that these settings show SASs performance in a more realistic light. Specifically, we find that rating SASs on the chatbot data showed more bias compared to the synthetic data, and round-tripping using Spanish and Danish as intermediate languages reduces the bias (up to 68% reduction) in human-generated data while, in synthetic data, it takes a surprising turn by increasing the bias! Our findings will help researchers and practitioners refine their SAS testing strategies and foster trust as SASs are considered part of more mission-critical applications for global use.
Representative Publications
- Kausik Lakkaraju, Aniket Gupta, Biplav Srivastava, Marco Valtorta and Dezhi Wu. The Effect of Human v/s Synthetic Test Data and Round-tripping on Assessment of Sentiment Analysis Systems for Bias. IEEE TPS 2023.
[Paper] [GitHub Respository] [BibTex]
Advances in Automatically Rating the Trustworthiness of Text Processing Services
Collaborators: AIISC, Department of Computer Science and Engineering - University of South Carolina, IBM Research
Contributors: Biplav Srivastava, Kausik Lakkaraju, Mariana Bernagozzi, Marco Valtorta
AI services are known to have unstable behavior when subjected to changes in data, models or users. Such behaviors, whether triggered by omission or commission, lead to trust issues when AI works with humans. The current approach of assessing AI services in a black box setting, where the consumer does not have access to the AI's source code or training data, is limited. The consumer has to rely on the AI developer's documentation and trust that the system has been built as stated. Further, if the AI consumer reuses the service to build other services which they sell to their customers, the consumer is at the risk of the service providers (both data and model providers). Our approach, in this context, is inspired by the success of nutritional labeling in food industry to promote health and seeks to assess and rate AI services for trust from the perspective of an independent stakeholder. The ratings become a means to communicate the behavior of AI systems so that the consumer is informed about the risks and can make an informed decision. In this paper, we will first describe recent progress in developing rating methods for text-based machine translator AI services that have been found promising with user studies. Then, we will outline challenges and vision for a principled, multi-modal, causality-based rating methodologies and its implication for decision-support in real-world scenarios like health and food recommendation.
Representative Publications
Why is my System Biased?: Rating of AI Systems through a Causal Lens
Collaborators: AIISC - University of South Carolina
Contributors: Kausik Lakkaraju
Artificial Intelligence (AI) systems like facial recognition systems and sentiment analyzers are known to exhibit model uncertainty which can be perceived as algorithmic bias in most cases. The aim of my Ph.D. is to examine and control the bias present in these AI systems by establishing causal relationships and also assigning a rating to these systems, which helps the user to make an informed selection when choosing from different systems for their application.
Representative Publications
ROSE: Tool and Data ResOurces to Explore the Instability of SEntiment Analysis Systems
Collaborators: AIISC, Department of Computer Science and Engineering - IIIT Naya Raipur
Contributors: Gaurav Mundada, Kausik Lakkaraju, & Biplav Srivastava
Sentiment Analysis Systems (SASs) are data-driven Artificial Intelligence (AI) systems that assign a score conveying the sentiment and emotion intensity when a piece of text is given as input. Like other AI, and especially machine learning (ML) based systems, they have also exhibited instability in their values when inputs are perturbed with respect to gender and race, which can be interpreted as biased behavior. In this demonstration paper, we present ROSE, a resource for understanding the behavior of SAS systems with respect to gender. It consists of data consisting of input text and output sentiment scores and a visualization tool to explore the behavior of SAS. We calculated the output sentiment scores using off-the-shelf SASs and our deep-learning-based implementations based on published architectures. ROSE, created using the d3.js framework, is publicly available here for easy access.
Representative Publications
- MUNDADA, G., LAKKARAJU, K., & SRIVASTAVA, B. ROSE: Tool and Data ResOurces to Explore the Instability of SEntiment Analysis Systems.
[Paper] [GitHub Respository] [BibTex]