Research Projects

Origin of the Proposal India is moving towards achieving its goals that are set up for Viksit Bharat [1]. To ground down the goals, the Govt. of India is taking several initiatives [1]. One of the important tasks for the government or the organisation is to maximize the coverage for their policy or maximize the influence of their action. For example, motivating people to get vaccinated. Further, assisting people with medical and basic facilities in emergency situations like the Manipur violence. Similarly, many applications like Emergency Response [2], Viral marketing [3], Product Adoption [4], Promote Public Health [5], Political Campaign [6], Advertisement [7], also aims to maximize the coverage in respect of their applications. There are several challenges that arise in maximizing the coverage. It includes the issue of reachability: not all nodes have a direct connection, unattainability: not everyone has the power to influence, and limited budget: there is limited budget to maximize the coverage. Influence maximization (IM) is a powerful computational problem that maximize the coverage of information, behaviour, or innovation. Existing algorithms suggest strategies, but they assume homogeneous propagation without accounting for underlying community polarization, belief-based fltering, etc. Therefore, existing algorithms are not applicable in real-world applications, especially in India, which is rich in diverse communities on the basis of genders, languages, castes, states, and cultures. To address the above limitation, this proposal aim to fll the gap by addressing the following objectives. Addressing the computational social science approach to maximize the coverage: Ignoring belief similarity or dissimilarity among users often results in overestimated or misplaced coverage in the network. To address this, our proposal will take the parameters like belief, polarization in communities, and application-oriented modelling into consideration. How to check the robustness of an approach in the real world: Existing algorithms aim to maximize the influence or coverage, but existing work failed to check the robustness in different cases. Therefore, this project will suggest an approach that will check the robustness of the algorithm considering position and negative sentiments. Integration of AI and NLP to maximize the coverage based on empirical data: Existing algorithms use a heuristic-based approach to estimate the parameters. However, to maximize the coverage in the real world, it requires an accurate estimation of these parameters. This project will estimate these parameters and future key nodes based on empirical data from the real world. Interpretability in IM: Existing work are lack interpretable diffusion explanations such as ”who influenced whom,” which is essential for many applications. This work aims to fll the gap by suggesting algorithms or diffusion models that are interpretable in IM.