Ajay Vishwanath

Affiliated with UiA Department of Information and Communication Technology

Email: ajay.vishwanath@uia.no
Office: A2090 (Jon Lilletuns vei 9, 4879 Grimstad, Norway)

Press photo Download business card

Background and experience

3 years of experience in the IT industry
3 years in artificial intelligence research with specialization in human robot interaction, natural language processing, and affective computing.

Areas of responsibility

I come from a tech background with diverse interests, ranging from information technology to computer vision. After my Masters specialization in artificial intelligence at National University of Singapore, I became interested in affective computing and AI ethics. My current research is based on artificial morality, where I focus on developing and training an artificial agent to be virtuous.

Research

Artificial Morality (aka Machine Ethics, AI alignment), virtue ethics, Explainable AI, AI and Ethics

Selected publications

Gupta, R. K., Vishwanath, A., & Yang, Y. (2020). COVID-19 Twitter dataset with latent topics, sentiments and emotions attributes. arXiv e-prints, arXiv-2007.

Vishwanath, A., Singh, A., Chua, Y. H. V., Dauwels, J., & Magnenat-Thalmann, N. (2019, October). Humanoid co-workers: How is it like to work with a robot?. In 2019 28th IEEE International Conference on Robot and Human Interactive Communication (RO-MAN) (pp. 1-6). IEEE.

Baka, E., Vishwanath, A., Mishra, N., Vleioras, G., & Thalmann, N. M. (2019, June). “Am I talking to a human or a robot?”: A preliminary study of human’s perception in human-humanoid interaction and its effects in cognitive and emotional states. In Computer Graphics International Conference (pp. 240-252). Springer, Cham.

Vishwanath, Ajay & Slavkovik, Marija (2025). Machine Ethics or AI Alignment? CEUR Workshop Proceedings. 3975, p. 81–90. doi: https:/ceur-ws.org/Vol-3975/paper9.pdf. Full text in Research Archive Show summary
Machine ethics and AI alignment have become increasingly important research disciplines in the context of modern AI technologies. Recently, there has been ambiguity regarding the specific aspects of ethical AI development that pertain to the two academic disciplines, often resulting in the interchangeable use of these terminologies. In this position paper, we explore these two disciplines by discussing ongoing research and clarifying their goals. Machine ethics researchers aim to embed ethical theories within AI, while AI alignment researchers develop techniques to ensure that AI agents behave as intended by humans. Based on these insights, we highlight overlaps, drawbacks, and motivate further interdisciplinary collaborations.
Vishwanath, Ajay & Omlin, Christian Walter Peter (2024). Exploring Affinity-Based Reinforcement Learning for Designing Artificial Virtuous Agents in Stochastic Environments. In Farmanbar, Mina; Tzamtzi, Maria P.; Verma, Ajit Kumar & Chakravorty, Antorweep (Ed.), Frontiers of Artificial Intelligence, Ethics, and Multidisciplinary Applications. Springer. ISSN 9789819998357. p. 25–38. doi: 10.1007/978-981-99-9836-4_3. Full text in Research Archive Show summary
Artificial virtuous agents are artificial intelligence agents capable of virtuous behavior. Virtues are defined as an excellence in moral character, for example, compassion, honesty, etc. Developing virtues in AI comes under the umbrella of machine ethics research, which aims to embed ethical theories into artificial intelligence systems. We have recently suggested the use of affinity-based reinforcement learning to impart virtuous behavior. Such a technique uses policy regularization on reinforcement learning algorithms, and it has advantages such as interpretability and convergence properties. Hence, we evaluate the efficacy of affinity-based reinforcement learning to design artificial virtuous agents using a stochastic role-playing game environment. Our results show that virtuous behavior can indeed result in our Papers, Please environment, and that algorithmic convergence can be controlled by the relevant hyperparameters. We then discuss some insights from our empirical evaluation of this method and motivate future research directions.
Vishwanath, Ajay & Omlin, Christian Walter Peter (2024). Localized Affinity-Based Reinforcement Learning for Interpretable State-Specific Decision-Making. In Bramer, Max & Stahl, Frederic (Ed.), Artificial Intelligence XLI: 44th SGAI International Conference on Artificial Intelligence, AI 2024. Springer. ISSN 9783031779145. p. 221–234. doi: https:/link.springer.com/chapter/10.1007/978-3-031-77915-2_16. Full text in Research Archive
Vishwanath, Ajay; Bøhn, Einar Duenger; Granmo, Ole-Christoffer; Maree, Charl & Omlin, Christian Walter Peter (2022). Towards artificial virtuous agents: games, dilemmas and machine learning. AI and Ethics. ISSN 2730-5953. doi: 10.1007/s43681-022-00251-8. Full text in Research Archive

View all works in NVA

Vishwanath, Ajay (2024). Beyond Algorithms: The Quest for Artificial Morality. Full text in Research Archive
Vishwanath, Ajay (2024). Macintyrean AI: A Functional Framework to Develop Virtuous Artificial Intelligence. Full text in Research Archive
Vishwanath, Ajay & Omlin, Christian Walter Peter (2024). Localized Affinity-based Reinforcement Learning for Interpretable State-specific Decision-making. Full text in Research Archive
Vishwanath, Ajay & Omlin, Christian Walter Peter (2024). Localized Affinity-based Reinforcement Learning. Full text in Research Archive
Vishwanath, Ajay & Omlin, Christian Walter Peter (2023). Exploring Affinity-based Reinforcement Learning for Designing Artificial Virtuous Agents in Stochastic Environments. Full text in Research Archive
Vishwanath, Ajay (2025). Engineering Artificial Morality: Virtues, Reinforcement Learning, and Game Environments. Universitetet i Agder. ISSN 9788284272900. Full text in Research Archive Show summary
Maskinetikk er et forskningsfelt som har som mål å inkludere etiske teorier i kunstig intelligens (AI), enten med et mål om å utvikle AI med kapasitet til å ta moralske prisverdige beslutninger, for å forhindre umoralske utfall, eller begge deler. En stor etisk teori, kjent som dydsetikk, er en normativ etisk teori som er forankret i dyder snarere enn normene, for å resultere i gode resultater uten å vektlegge selve resultatene. Dyder, er karakterisert som eksemplariske moralske trekk, og har hatt begrensede implementeringer innen maskinetikk, først og fremst på grunn av de iboende vanskelighetene med det ingeniørmessige aspektet av moralsk karakter. Dydsetikk samt kunstig intelligens nødvendiggjør, at en agent streber etter dyder og kontinuerlig søker å forbedre deres moralske karakter ved å lære av tidligere feil. Selv om vi ikke sidestiller menneskelige dyder med kunstige dyder, argumenterer vi for bruken av Alasdair MacIntyres dydsetiske rammeverk – forankret i sosiale praksiser, fortellinger og moralske tradisjoner – som grunnlag for å utvikle en filosofi om konstruksjon av dydig AI. Dette livslange læringsaspektet hos dydige agenter implementeres ved hjelp av forsterkende læring (RL). Imidlertid bruker tradisjonell RL hovedsakelig en mekanisme som utelukkende er rettet mot maksimering av belønninger. For å overvinne denne avhengigheten, tar vi i bruk et nytt RL-paradigm kjent som affinitetsbasert RL, som bruker policyregularisering for å definere prior som kan være en god rollemodell for en agent. Vi utvikler RL-miljøer basert på rollespill med moralske dilemmaer som Papers, Please og Fog of Love. Funnene fra begge eksperimentelle spillmiljøene, gjennom bruken av affinitetsbasert RL, fremhever at det er mulig å konstruere dydig oppførsel innen kunstig intelligens, som har betydelige implikasjoner for disiplinen maskinetikk.

View all works in NVA

Published Nov. 18, 2025 5:53 AM - Last modified Nov. 18, 2025 5:53 AM