Balaraman, Vivek Chief Scientist, Human Centric Systems Research Group, – Pune (TRDDC)
Balaraman, Vivek Chief Scientist, Human Centric Systems Research Group, TCS Innovation Labs – Pune (TRDDC) Recent Publications 1. Simulating Daily Routines to Estimate Convenience Store Footfalls Vivek Balaraman, Meghendra Singh, Deepa Athle Convenience stores or c-stores are medium sized retail establishments. C-stores compete not on price but on factors such as faster shopping, longer hours and proximity to home or work. Two key challenges in the cstore business are to decide on the best location given a set of potential locations, and how various convenience inducing factors such as proximity or any time availability can enhance footfall. Both challenges are tied to the problem of estimating footfall in a store. Studies across domains suggest that the daily routines of people are related to their shopping processes. Routines however have been significantly underresearched, both in general psychological studies and in studies of consumer behavior. A reason could be that trends in economic or consumer behavior are observable only at a macro-level, while fluctuating significantly at the micro level. There has also been a lack of necessary methodological tools to relate individual behavior in the micro to that of the macro. However, the birth of new computational techniques, such as agent based simulation enables us to derive emergent consumer behavior at a macro level, by simulating micro-level populations of consumers. In this paper, we first address the relevance of behavioral routines to consumer behavior, particularly to footfalls in stores. We then demonstrate how simulating the routines of a community using an agent based approach can help us forecast footfalls in c-stores. We then show how these, coupled with various convenience inducing factors, might be used to decide on store location and the growth of footfalls. 2. Human Behavioral Modeling for Enhanced Software Project Management Sandeep Athavale, Vivek Balaraman Despite advances in technology and the growing maturity of processes, software project outcomes are still highly dependent on the performance of individuals in the software project team. The individual performance is based on human aspects – cognitive, psychological and sociological apart from enablers such as infrastructure and environment. Project managers need tools that account for human aspects for improved software project planning and management. It will enable them to take preventive actions or make interventions to reduce the uncertainty of software project outcomes and increase the likelihood of project success. Human behavioral modeling and simulation can provide precisely such decision support. In this paper, we develop a computational model of human behavior in the context of a software development project. We model individual and interpersonal human aspects that affect task performance. Towards this end, we use the agent based modeling and simulation (ABMS) paradigm as it captures how each individual project member acts or performs in autonomous ways driven by their dispositions and how these autonomous members interact with each to produce project outcomes. It thus provides project managers with a powerful tool to experiment with further what-if scenarios. Through our model and prototype simulation, we show how the inclusion of human aspects leads to discovery of multiple possible project outcomes not visible through linear projections. 3. Using agent based simulation as a visual sandbox to help businesses draft customer acquisition strategy Vivek Balaraman Most businesses, commercial or social need a customer growth strategy. Dignity Foundation (DF) is a non-profit organization that provides useful support services to the elder segment (50+). DF are in the process of planning and strategizing their customer growth roadmap. DF have on the table a spectrum of member acquisition techniques they could use. They also need to consider customer loss due to nonrenewals. Complexity arises because of the interplay between individual member growth techniques which also differ in their modes of member growth. This complexity made it difficult for DF to plan and strategize the roadmap that would lead them to break even. DF wanted a sandbox that would allow them to realistically experiment with different member acquisition approaches. Critical requirements were that the sandbox should provide a visual feel, for the dynamics of member growth under different technique mixes. DF were not looking at econometric prediction approaches but at ways that would help them in creatively thinking out the possibilities. In this paper we present an Agent Based Simulation (ABS) approach to this problem. ABS enables us to simulate and visually depict a virtual population of DF's target segment, consisting of both subscribers and non-subscribers. We show how ABS helps us obtain a visual picture of various strategies playing out, their relative efficacy and how these impact subscriber base. We discuss the insights provided by the exercise. We conclude this paper with a look at the extensions to this work. 4. The Lokpal Debate: An Agent Based Simulation Perspective Vivek Balaraman Corruption in governance, particularly public institutions and in decisions taken by public funded institutions is an over-riding concern within many countries. A fierce debate has raged within India on the setting up of a Lokpal, an autonomous agency to which citizens can appeal when they face or observe corruption in local / regional / national governance institutions and bodies receiving public funding. But will the Lokpal be effective in reducing or mitigating corruption and in correcting wrong? In this paper we use agent based simulation (ABS) to understand under what circumstances a simple model of the Lokpal agency will be effective. We define three metrics that characterize the rightness of the decision being taken. We then discuss the results of simulating an environment with and without a Lokpal under various scenarios. For these scenarios we look at the performance on the three metrics as well as other measures to get a sense of the Lokpal’s effectiveness under various conditions. Together these experiments lead us to certain learnings. We conclude with a look at how this model may be extended.