In recent years, designing fairness-aware methods has received much attention in various domains, including machine learning, natural language processing, and information retrieval. However, understanding structural bias and inequalities in social networks and designing fairness-aware methods for various research problems in social network analysis (SNA) have not received much attention. In this project, we highlight how the structural bias of social networks impacts the fairness of different SNA methods. We define fairness aspects and evaluation metrics that should be considered while proposing network structure-based solutions for different SNA problems, such as link prediction, influence maximization, centrality ranking, and community detection. Our recent survey-cum-vision paper, titled "FairSNA: Algorithmic Fairness in Social Network Analysis," discusses it in further detail and highlights various open research directions that require researchers' attention to bridge the gap between fairness and SNA.
The Privacy Lost project uses web tracking (based on cookies and scripts embedded in websites) as a key exemplar of digital surveillance. The main aim of the project is to understand how the dynamics of tracking have been shaped by variations in economic, social and cultural factors across the world, and how this in turn sheds light on the development of surveillance capitalism. We analyze the historical data of web pages from multiple geographic and cultural regionsusing network science, information retrieval, and natural language processing, to highlight the core elements and dynamics of digital surveillance visible.
What is the potential of a 21st century learning environment that mirrors the capabilities of personalized Apps? In contrast to the standard linear or tree-like educational system of sequential lectures or chapters, we design a real-time, modular, adaptive teaching-learning environment for enhanced and personalized education, called the Curated Heuristic Using a Network of Knowledge (CHUNK) Learning concept. The CHUNK Learning model breaks away from the predictable pattern of traditional education models and provides content delivery that adapts to the capabilities, learning styles, and approaches to problem-solving for every learner. The CHUNK Learning tool is a student-centered teaching-learning system whose purpose is to make learning engaged, flexible, and respectful of the students' time. This system converts curricula into a network composed of nodes, where each node is referred to as a CHUNK of educational content and edges capture relationships that exist among the nodes. The Network of Knowledge is composed of lesson materials joined together by prerequisite relationships and common attributes based on competency or skill levels. Our main goal is how we can develop a personalized, streamlined path for learners engaged in e-learning to make interdisciplinary learning easy for students belonging to different fields. We are also working on a recommender system to recommend the most relevant CHUNKs to a learner based on the last CHUNK that they have completed and their learning goals.
It has been observed in several works that the ranking of candidates based on their score can be biased for candidates belonging to the minority community. In recent works, the fairness-aware representative ranking was proposed for computing fairness-aware re-ranking of results. The proposed algorithm achieves the desired distribution of top-ranked results with respect to one or more protected attributes. In this project, we highlight the bias in fairness-aware representative ranking for an individual and for a group if the group is sub-active on the platform. We define individual unfairness and group unfairness from two different perspectives. We aim to propose methods to generate ideal individual and group fair representative ranking if the universal representation ratio is known.
Today each online social network hosts millions of user accounts. These social networks provide an easy platform to share the information where most of the information is shared as a microblog. Due to the easy sharing of information, the spread of fake news and rumors has been prevalent. We have seen the impact of spreading of fake news on major events like the US election, Jakarta election, or distorting the reputation of a company. In this project, we focus on proposing the mitigation techniques using network science-based approach. We also reviewed the existing fake news detection and mitigation techniques, and what kind of actions can be taken further to control the fake news spreading.
In real-world complex networks, the importance of a node depends on two important parameters: 1. characteristics of the node, and 2. the context of the given application. The current literature contains several centrality measures that have been defined to measure the importance of a node based on the given application requirements. In this project, we aim to propose fast and efficient methods to estimate the global centrality rank of a node without computing the centrality value of all the nodes. These methods are further extended to estimate the rank without having the entire network. The proposed methods are based on the structural behavior of centrality measures, network properties, and sampling techniques. We have proposed methods to estimate the degree, closeness, and k-shell rank of the nodes. My Ph.D. thesis is based on this project.
Being as old as human civilization, discrimination based on various grounds as race, creed, gender, and caste is prevailing in the world from a long time. To undo the impact of this long-enduring historical discrimination, governments worldwide have adopted various forms of affirmative action; such as positive discrimination, employment equity, and quota system. Locally known as “Reservation” policy, affirmative action in India is one of the world's oldest and most complex affirmative policy. Although being one of the most controversial and frequently debated issues, the reservation system in India lacks a rigorous scientific study and analysis. In this paper, we discuss the dynamics of the reservation system based on the cultural divide among Indian population using social network analysis. The mathematical model, using Erdos-Renyi network, shows that the addition of weak ties between the two components leads to a logarithmic reduction in the social distance. Our experimental simulations establish the claim for the different clans of frequently studied social network models as well as real-world networks. We further show that a small number of links created by the process of reservation are adequate for a society to live in harmony.
For the past several decades, gender-based biases have been prevalent in society. The gender ratio bias is also present among the people working in STEM. For the past few decades, several govt, organizations, private institutions, NGOs, etc., have worked extensively to reduce the gender gap in STEM. Still, we are far from reaching an equal gender ratio. In this project, we focus on the analysis and modeling of biases in the communities based on gender, race, etc. and the impact of affirmative actions to remove these biases from society. We work towards modeling of biases, actions for removing them, influence propagation, and the emergence of leaders in such scenarios. Apart from this, we also study gender bias in online learning platforms. How different people acquire different roles for the stability of the ecosystem and how they converge over time. This project opens a wide range of questions that are yet to be explored and answered.
Decision makers use partial information networks to guide their decision, yet when they act, they act in the real network or the ground truth. Therefore, a way of comparing the partial information to ground truth is required. In this project, we aim to propose methods for comparing the evolution process of real-world networks. We introduced a statistical measure that analyzes the network obtained from the partially observed information and ground truth, which of course can be applied to the comparison of any networks. As a first step, in the current research, we restrict ourselves to networks of the same size to introduce such a method, which can be generalized to different size networks. Next, we focus on generalizing the proposed method to compare the network of different sizes.
Real-world scale-free networks possess both the community as well as the core-periphery meso-scale structures that shows the modular and hierarchical organization in the given networks. This project mainly focuses on understanding the evolving phenomenon and coexistence of core-periphery community structures. Based on our observations, we further propose evolving models to generate synthetic weighted and unweighted scale-free networks having both meso-scale structures.
In this project, we present a synthesized analysis of three terrorist networks through the analysis of the multiple layers of these networks. First, we study how these networks have different characteristics than scale-free real-world networks. The main challenges associated with these networks are incompleteness, fuzzy boundaries, and dynamic behavior. We account these characteristics and propose a method to identify knowledge sharing communities (KSC). We also proposed models to generate multilayered synthetic networks having similar properties.
For network scientists, it has always been an interesting problem to identify the influential nodes in a given network. K-shell decomposition method is a widely used method which assigns a shell-index value to each node based on its influential power. K-shell method requires the entire network to compute the shell-index of a node that is infeasible for large-scale real-world dynamic networks. In this project, we focus on estimating the shell-index of a node using local neighborhood information. Next, we use the estimated coreness value to estimate the global coreness rank of a node without having the entire network.
In this project, we study how does a meme spread on the network. We study real-world meme spreading datasets to observe the role of core nodes in making a meme viral. We further study the impact of core-periphery and community structure on the meme spreading. Based on our observations, we proposed a meme spreading model using penta-level classification of edges in the network. The proposed spreading model is verified using Twitter datasets.
Assistant Professor | May 2023 to Present | Netherlands
Research Fellow | February 2023 to April 2023 | Netherlands
Research Fellow | March 2020 to February 2023 | Netherlands
Postdoctoral Fellow | November 2018 to March 2020 | Singapore
Research Scholar | July 2014 to October 2018 | Punjab, India
IT Officer (Assistant Manager) | August 2013 to July 2014 | Patiala, Punjab, India
Software Developer | July 2011 to August 2013 | New Delhi, India
Summer Intern | May 2010 to July 2010 | New Delhi India