Nabiha
Asghar,
PhD
candidate
David
R.
Cheriton
School
of
Computer
Science
In human-computer interaction (HCI), one of the technological goals is to build human-like artificial agents that can think, decide and behave like humans during the interaction. A prime example is a dialogue system, where the agent should converse fluently and coherently with a user and connect with them emotionally. Humanness and emotion-awareness of interactive artificial agents have been shown to improve user experience and help attain application-specific goals more quickly. However, achieving human-likeness in HCI systems is contingent on addressing several philosophical and scientific challenges. In this thesis, I address two such challenges: replicating the human ability to 1) correctly perceive and adopt emotions, and 2) communicate effectively through language.
Several research studies in neuroscience, economics, psychology and sociology show that both language and emotional reasoning are essential to the human cognitive deliberation process. These studies establish that any human-like AI should necessarily be equipped with adequate emotional and linguistic cognizance. To this end, I explore the following research directions.
- I study how agents can reason emotionally in various human-interactive settings for decision-making. I use Bayesian Affect Control Theory, a probabilistic model of human-human affective interactions, to build a decision-theoretic reasoning algorithm about affect. This approach is validated on several applications: two-person social dilemma games, an assistive healthcare device, and robot navigation.
- I develop several techniques to understand and generate emotions/affect in language. The proposed methods include affect-based feature augmentation of neural conversational models, training regularization using affective objectives, and affectively diverse sequential inference.
- I devise an active learning technique that elicits user feedback to produce natural and coherent language during a conversation.
- I explore incremental domain adaptation in language classification and generation models. The proposed method seeks to replicate the human ability to continually learn from new environments without forgetting old experiences.