Research Interests and Activities
My current research interests are in AI, BPM and workflow systems, Blockchain technology, Business analytics, Deep learning, Health IT and Healthcare forums, and process mining. I have collaborated extensively with researchers in US, and internationally with many co-authors at the Technical University of Eindhoven, Netherlands; Ulm University, Germany; Tsinghua University, China; Chinese University of Hongkong; and Queensland University of Technology, Brisbane.
My recent work with Om Dobariya on how the tone of prompts affects accuracy of ChatGPT results has received widespread attention in media. Om Dobariya and Akhil Kumar. "Mind Your Tone: Investigating How Prompt Politeness Affects LLM Accuracy" (arxiv. org/abs/2510.04950)
Currently, I am very excited about the opportunities presented by GenAI. I never imagined that GenAI would make as much progress in my lifetime as it has. I started working in AI in the late 1980s as a faculty member at Cornell Johnson School of Management. I was teaching MBAs about Expert Systems at that time. I also wrote a paper on neural networks with Vithala Rao. AI has gone through so many winters and springs over the last 50+ year. But what we see today... No way! it is from SciFi!
In our research cited above we are investigating how LLMs like ChatGPT respond to different wordings of the same prompt. How does it affect accuracy on multiple choice questions. We find that LLMs are not always consistent in their responses. Minor cues in the prompt (firm, curt, rude, polite, nice, etc.) can affect the answers from LLMs, even for multiple choice questions where the answers are precise (A, B, C, D only). Thus, protocols for human-AI conversation are different from those for human-human conversation. For the longest of times, we humans have wanted conversational interfaces for interacting with machines. But now we realize that there are drawbacks for such interfaces too and there is some value in APIs that are structured. We are looking into ways that prompts can be optimized so that they produce consistent results. Another important factor is the energy consumed in a running a query or a prompt. It is important to consider how any prompt can be optimized for energy use. Other topics that we are exploring are process optimization using LLMs and system design techniques using LLMs. In all these topics consistency, reliability and repeatability of answers is very important.
I serve as an associate editor for ACM Transactions on Management Information Systems. I was previously an associate editor for IEEE Transactions on Services Computing, INFORMS Journal on Computing and Information Systems Research. I was a co-program-chair of BPM 2017 Conference. I also served as a co-program chair of CoopIS'11 and as a co-chair of Workshop on Information Technologies and Systems (WITS'07). I was a principal investigator for National Science Foundation, and also received support from IBM, Sun Microsystems and other organizations for his work.