Swati Gupta

I joined Callaghan Innovation in November 2014. Prior to that I was a Scientist at the Computational Social Cognition group of IHPC, A*STAR in Singapore for six years. At A*STAR, I worked on a variety of research projects in the areas of communication theory, social psychology, and cognitive science. I did my PhD at the University of Sheffield, UK, in Natural Language Generation for Spoken Dialogue Systems.

My research focus is Human-Computer Interaction, inspired by theories in Cognitive Science, Psychology and Linguistics. I am deeply interested in applying my research to social, environmental and health problems. 

I am actively involved with several projects listed on this website. Below are some of my other ongoing projects that are not part of the Assistive Technologies team.

Deception as a Social Strategy

Verbal Deception is an intentional verbal act of controlling information to prevent the hearer (H) from believing what the speaker (S) believes is true, and to make H believe what S believes is not true. Deception can be strategic when truth conflicts with personal or social, selfless or selfish goals. People tend to select one deception type rather than others depending on the situation they’re are and who they’re talking to. They do so in order to manipulate some beliefs of the hearer. The objective of this project is to understand the underlying process of deceptive strategy selection by developing a theoretical model of the interaction between various deception types and the belief-manipulation goals of deception.

Such a model provides a basis for building socially intelligent systems which could be applied in domains such as elder care, tutoring, and corporate training.

Collaborators: Andrew Ortony, Kayo Sakamoto

MoCHA: Monitoring Cognitive Health using Apps

Many countries are experiencing rapidly-aging populations, and care for Alzheimer’s Disease (AD) is going to impose a huge burden. Therapies for AD are only effective if it’s detected early, but the state of the art is paper-based exams in doctor offices, which is not only very stressful, people also often avoid is because of the fear of a potential social stigma. We created a set of tablet-based “serious games” that monitor for the same kind of signs as the typical exams, but the games are fun and are meant to be enjoyed by elder players long before any signs of impairment appear.

This has the potential to detect AD earlier and more cost-effectively. This framework can also be applied to asses general cognitive health as well as other dementia related problems.

Collaborators: David Pautler, Ilya Farber, Karl Fua, Angus Gellatly, Joseph Simons, Kumaresh Darmalingam, Keith Tan, Terato

Past Projects

Choice Blindness

Choice blindness is a psychological phenomenon which was first described in 2005 in Science. The experimental paradigm is based on a simple choice task, in which a subject is shown two cards with photos of faces and asked to point to the one which they find more attractive. Seconds later, the subject is shown their chosen card again, and asked to explain what features of that face motivated them to choose it. Unbeknownst to the subjects, on a small number of trials a simple card-manipulation trick is used to show them the card that they did not choose. The very surprising result was that subjects not only usually failed to notice the switch, but then went on to offer detailed, confident, extended explanations for a choice which was the opposite of the one that they actually made.

There are two features that make choice blindness potentially very significant. First, it is one of the very few paradigms which are able to reliably produce extended confabulation (sincere false report) in normal subjects, and the only one which is able to do so in a single session; normally this sort of confabulation is seen only in patients with severe neuropsychiatric disorders or brain lesions. Choice blindness thus offers a window on a behavior that is otherwise very difficult to study. Second, choice blindness has deep implications for theories of decision-making and self-understanding. One interpretation is that subjects do not actually remember the reasons for their choices even in the non-manipulated trials: in other words, people don’t actually have any memory for the choice process in most cases, and instead “make up” explanations as needed. If true, this could have very significant implications for a number of practical domains, including law (“intent” and witness testimony), marketing (focus groups), group decision-making and introspective therapy.

We conducted a study which closely replicates the original face-choice experiment (albeit computerized and using an Asian cohort), with a few key modifications and also investigate it’s effect on preference alteration.

Collaborators: Ilya Farber, Fumihiko Taya, O’Dhaniel A. Mullette-Gillman

Modelling Impression Formation

First impression formation is the process by which people create a schema of someone they meet for the very first time by observing and interpreting their behaviour (e.g. smiles a lot, talks politely), and the categories or social roles they belong to (e.g. teacher, female). This schema consists of their perception of the target in terms of the kind of person they think he/she is. Impression formation helps in understanding the nature of the target person better and making predictions about their future behavior. It is crucial for any kind of social interaction; ranging from recruitment to speed dating scenario.

We are particularly interested in exploring: a) if it is possible to design a single theoretical, and based on that, a computational framework to simulate the generic impression formation process, and b) the nature and interaction of various socio-cognitive constructs, such as goals, beliefs, attitude, values, as well as the role of the specific context in the process of first impression formation.

Collaborators: Tei Laine, Brian Monroe, Andrew Ortony

Modelling Linguistic Politeness

Doctoral thesis. This research draws from a sociolinguistics theory of politeness and combines a Spoken Language Generator with an AI planner to model linguistic politeness in collaborative task oriented dialogue with the ultimate goal of providing a stimulating environment and a fun way for learning politeness in English as a Second Language (ESL). Embodied Conversational Agents demonstrate the dialogues to the human ESL learners and interact with them using spoken language to practice politeness in role play situations.

The AI Planning mechanism, given a goal, generates steps to achieve the goal. The Spoken Language Generator generates goals to communicate about the Plan steps via speech acts like request, offer, etc, extracts components like verb, subject and object from the preconditions, steps and effects of the Plan and creates multiple variations for different politeness strategies for an utterance. It then runs an optimization algorithm and extracts the right politeness strategy from this repository to generate polite dialogues as per the given situation. These dialogues along with their recognition grammar, which is also automatically generated using similar generative mechanism, are sent to the Embodied Agents who interact with the learners.

The project also involved conducting studies to verify the politeness strategies used and to test the pedagogical objectives.

Advisors: Marilyn Walker, Daniela Romano

Aligning Hindi and Urdu Bilingual Corpora for Projecting POS-tags

Masters Project. This project aimed to correlate two popular South Asian languages: Hindi and Urdu. The task involved aligning their parallel corpora. The technique used for sentence alignment was based on sentence length in terms of number of words in the sentence. Technique for word alignment was based on cognates. It further investigated the possibility of Projecting POS-tags across from the POS-tagged Urdu corpus to the un-annotated Hindi corpus and demonstrated the feasibility of training a stand alone POS-tagger for Hindi language. The corpus used was the EMILLE corpus – Enabling Minority Language Engineering, developed primarily by the Lancaster University, UK in collaboration with other institutions.

Advisor: Mark Hepple

Publications

Monroe, B.M., Laine T., Gupta S., and Farber I. (2017). Using connectionist models to capture the distinctive psychological structure of impression formation. In Vallacher, R.R.,  Nowak, A.,  Read, S.J. (Eds.), Computational Social Psychology (Frontiers of Social Psychology). Routledge Publication.

Farber, I., Fua, K.C., Gupta, S., Pautler, D. (2016). MoCHA: Designing Games to Monitor Cognitive Health in Elders at Risk for Alzheimer's Disease. Proceedings of the 13th International Conference on Advances in Computer Entertainment Technology (ACE). Osaka, Japan. Nov 09-12. ACM Publication. DOI:10.1145/3001773.3001818

Nunnerley, J., Gupta, S., Snell, D., King, M.J. (2016). Training wheelchair navigation in immersive virtual environments for patients with spinal cord injury – end-user input to design an effective system. Disability and Rehabilitation: Assistive Technology. DOI: 10.1080/17483107.2016.1176259

Graham, H., Bond, A., McCormick, M., Hobbs, O., Yoo, C., Gupta, S., Mulligan, H., & King, M. (2016). A novel communication application to encourage social interaction by children with autism spectrum disorder. New Zealand Journal of Physiotherapy, 44(1), 50-57. doi: 10.15619/NZJP/44.1.06

Ortony, A., Gupta, S. (2015). Strategies for Achieving the Goals of Deception. To appear In Meibauer, J. (Ed), The Oxford Handbook of Lying. Oxford: Oxford University Press

Taya, F., Gupta, S., Farber, I., and Mullette-Gillman, O.A. (2014). Manipulation Detection and Preference Alterations in a Choice Blindness Paradigm. PLoS ONE 9(9): e108515. doi:10.1371/journal.pone.0108515.

Fua, K., Gupta, S., Pautler, D., & Farber, I. (2013). Designing Serious Games for Elders. Proceedings of the 8th International Conference on the Foundations of Digital Games (FDG 2013), pp. 291-297. May 14-17, 2013. Chania, Crete, Greece.

Gupta, S., Sakamoto, K. (2014). Equivocation (4 pages). In Levine, Timothy R. & Golson, J. Geoffrey (Eds.), Encyclopedia of Lying and Deception. SAGE publication.

Gupta, S., Sakamoto, K. (2014). Half-truth (4 pages). In Levine, Timothy R. & Golson, J. Geoffrey (Eds.), Encyclopedia of Deception. SAGE publication.

Gupta, S., Sakamoto, K. (2014). Definition: Deception (3 pages). In Levine, Timothy R. & Golson, J. Geoffrey (Eds.), Encyclopedia of Lying and Deception. SAGE publication.

Sakamoto, K., Gupta, S. (2014). Rationality (4 pages). In Levine, Timothy R. & Golson, J. Geoffrey (Eds.), Encyclopedia of Lying and Deception. SAGE publication.

Sakamoto, K., Gupta, S. (2014). Cost of lying (4 pages). In Levine, Timothy R. & Golson, J. Geoffrey (Eds.), Encyclopedia of Lying and Deception. SAGE publication.

Sakamoto, K., Gupta, S. (2014). Acceptability of lying (3 pages). In Levine, Timothy R. & Golson, J. Geoffrey (Eds.), Encyclopedia of Lying and Deception. SAGE publication.

Gupta, S. (2012). Review of the book Politeness in East Asia, by Kádár, Dániel Z. and Mills, Sara (eds). In: Language and Dialogue 2:2. 2012 iv, 145 pp. (pp. 313–316)

Gupta, S., Sakamoto, K., & Ortony, A. (2012). Telling it like it isn’t: a comprehensive approach to analyzing verbal deception. In F. Paglieri, L. Tummolini, R. Falcone & M. Miceli (Eds.), The goals of cognition: Festschfit for Cristiano Castelfranchi. London, College Publications.

Gupta, S., Walker, M.A., Romano, D.M. (2008). Using a Shared Representation to Generate Action and Social Language for a Virtual Dialogue Environment. Proceedings of the Association for the Advanced of Artificial Intelligence (AAAI 2008) Spring Symposium on Emotion, Personality and Social Behavior, Stanford University, Palo Alto, CA.

Gupta, S., Walker, M.A., Romano, D.M. (2008). POLLy: A Conversational System that uses a Shared Representation to Generate Action and Social Language (demo paper). Proceedings of the 3rd International Joint Conference on Natural Language Processing (IJCNLP 2008), 8-12 January 2008, Hyderabad, India.

Gupta, S., Walker, M.A., Romano, D.M. (2007). How Rude are You?: Evaluating Politeness and Affect in Interaction. Affective Computing & Intelligent Interaction (ACII-2007), 12th September 2007, Calouste Gulbenkian Foundation, Lisbon, Portugal. Lecture Notes in Computer Science (LNCS) Springer Berlin/Heidelberg, ISSN 0302-9743, vol. 4738, pp. 203-217. Intensity of Perceived Emotions in 3D Virtual Human.

Gupta, S., Walker M. A. (2007). Generating Politeness in Task Based Interaction: An Evaluation of the Effect of Linguistic Form and Culture.Proceedings of ENLG'07, the 11th European Workshop on Natural Language Generation, Germany, 17-20 June 2007.

Gupta, S., Romano, D.M., Walker M. A. (2005). Politeness and Variation in Synthetic Social Interaction. Proceedings of the H-ACI Human-Animated Characters Interaction Workshop, British HCI 2005, The 19th British HCI Group Annual Conference Napier University, Edinburgh, UK 5-9 September 2005.

Lu, Y., Lu, F., Sehgal, S., Gupta, S., Du, J., Tham, C. H., Green, P. D., Wan, V. (2004) Multitask Learning in Connectionist Speech Recognition, Proceedings of AICSSE - Tenth Australian International Conference on Speech science & Technology, 8 - 10 December 2004, Macquarie University, Sydney, Australia.

Posters and Presentations

Gupta, S. (2015). ReVIVE- Rehabilitation Via Immersive Virtual Environments. Talk at the Burwood Academy of Independent Living, Christchurch, New Zealand. 5 Nov 2015.

Gupta, S., King, M. (2015). Designing Games for Individuals with Autism Spectrum. Poster presented at i-CREATe 2015, Singapore, 11-14 Aug.

Gupta, S., King, M. (2015). Immersive virtual reality rehab post spinal cord injury. Demo presented at the 2015 Healthcare Congress, 16-17 June, Auckland, New Zealand.

Simons, J. J. P., Laine, T., Sakamoto, K., Gupta, S., Gellatly, A.R.H., Wan, K.S., & Farber, I. (2014). Differential attitudes to deception and risk: Validation of a self-report scale. Poster presented at the annual meeting of the Society for Personality and Social Psychology, Austin, TX.

Gupta, S., Laine, T., Sakamoto, K., Ortony, A., (2013). Deception as a social strategy. Paper presented at the workshop on Deception as a Social strategy, 35th Cognitive Science Conference, Berlin, Germany.

Farber, I., Taya, F., Gupta, S., Mullette-Gillman, O.A., (2013a). The temporal dynamics of choice blindness: flat detection rates and short-term preference alterations. Poster presented at the 35th Cognitive Science Conference, Berlin, Germany.

Farber, I., Taya, F., Gupta, S., and Mullette-Gillman, O.A., (2013b). The mechanism of choice blindness: clues from patterns of preference alteration. Paper presented at the Association for the Scientific Study of Consciousness (ASSC) 17 meeting, San Diego, USA.

Taya, F., Gupta, S., Farber, I., and Mullette-Gillman, O.A., (2013a). Preference alterations induced by choice belief. Poster presented at the Society for Neuroeconomics Conference, Switzerland.

Taya, F., Gupta, S., Farber, I., and Mullette-Gillman, O.A., (2013b). Choice Manipulation Alters Preferences: Effects of Time and Detection in a “Choice Blindness” Paradigm. Poster presented at the Singapore General Hospital Annual Scientific Meeting, Singapore.

Sakamoto, K., Laine, T., Gupta, S., & Ortony, A. (2012). Deceptive Strategy Selection as Decision Making Under Risk: A taxonomy and proposed experiment. Poster presented at the Deception, Incentive, and Behavior Symposium, San Diego, CA, April 2012.

Monroe, B.M., Laine, T., Gupta, S., Sadeghi, S. & Farber, I. (2012).  A Test of Predictive Validity of Connectionist Models of Person Judgments. Talk presented at Social Dynamics and Computational Modeling Preconference, Society of Personality and Social Psychology Annual Conference, January 2012, San Diego, USA.

Gupta, S., Laine, T., Monroe, B.M. (2010). Making a good impression (- formation model). Invited talk at Nanyang Technological University, Singapore.

Laine, T., Gupta, S., Monroe, B. (2010), "Making a good impression (-formation model). Poster presentation at the Annual Meeting of the Cognitive Science Society, Portland, Oregon.