Thao Le

(she/her)

Email  /  Google Scholar  /  DBLP  /  ORCID  /  Twitter  /  Github

I am a PhD Candidate in Explainable AI (XAI) in the AI and Autonomy Lab, School of Computing and Information Systems, The University of Melbourne, under the supervision of Tim Miller (primary supervisor), Liz Sonenberg and Ronal Singh. My research interests include explainable AI, human-computer interaction, human-centered decision support and human-AI planning.

My PhD research focuses on two themes: (1) explaining the uncertainty and (2) improving explainable decision support. First, I propose and evaluate a method for explaining model uncertainty and therefore, promote trust and improve understanding for users. Second, we propose a new decision-making model which gives promising directions in reducing under-reliance and over-reliance. The new framework is evaluated in various application domains, such as the economics field and skin cancer diagnosis.

I apply mixed methods in my research by using machine learning techniques, designing/conducting behavioural experiments and doing qualitative/quantitative data analysis. My skills include (1) Python, PyTorch, Scikit-learn, Pandas, (2) machine learning/reinforcement learning techniques, (3) high performance computing (HPC), (4) building web-based applications for human-AI interaction.

Selected Publications

[ECAI 2024] Towards the New XAI: A Hypothesis-Driven Approach to Decision Support Using Evidence
Thao Le, Tim Miller, Liz Sonenberg, Ronal Singh
European Conference on Artificial Intelligence (ECAI). Main Track. Acceptance Rate: 547/2,344 (23%).
Spain.
[Paper]

Describing and evaluating an approach for hypothesis-driven XAI based on the Weight of Evidence (WoE) framework.

Keywords: XAI, decision-making, hypothesis-driven, evidence, evaluative AI

[AAAI 2023] Explaining Model Confidence Using Counterfactuals
Thao Le, Tim Miller, Ronal Singh, Liz Sonenberg
Association for the Advancement of Artificial Intelligence (AAAI). Main Track. Acceptance Rate: 1,721/8,777 (19.6%).
Washington DC, USA.
[Paper]

Formalising counterfactual explanation of confidence and showing that this explanation can help users better understand and trust the model.

Keywords: uncertainty/confidence, counterfactual explanation, HCI

[JAIR 2021] Goal Recognition for Deceptive Human Agents through Planning and Gaze
Thao Le, Ronal Singh, Tim Miller
Journal of Artificial Intelligence Research.
[Paper]

Presenting new models for goal recognition under deception using a combination of gaze behaviour and observed movements of the agent.

Keywords: planning, HCI

Peer-Reviewed Workshops/Doctoral Consortia

  1. [AAAI DC 2023] Explaining the Uncertainty in AI-Assisted Decision Making
    Thao Le
    AAAI Doctoral Consortium.
    Washington DC, USA.
    [Paper]
  2. [XAI-IJCAI 2022] Improving Model Understanding and Trust with Counterfactual Explanations of Model Confidence
    Thao Le, Tim Miller, Ronal Singh, Liz Sonenberg
    International Joint Conference on Artificial Intelligence (IJCAI-ECAI). Workshop on Explainable Artificial Intelligence.
    Vienna, Austria.
    [Paper]

Academic/Research Experience

  • Academic Tutor (Teaching Assistant) - AI Planning for Autonomy, COMP90054 (Graduate level) - S1/S2 2023, S1/2024
    • Topics covered: Search algorithms and heuristic functions, classical (AI) planning, Markov Decision Processes, reinforcement learning, game theory
  • Research Assistant - AI and Autonomy Lab, The University of Melbourne - 05/2022
    • Project: Cognitive Models Supporting Human-Machine Partnerships

Academic Service

  • Reviewer: HCOMP 2023

Selected Awards/Achievements

  • Google Conference Grant (2023)
  • CIS Nominee for Apple Scholars in AI/ML PhD Fellowships (2022/2023)
  • Melbourne Centre for Data Science (MCDS) Top Up Scholarship (2022)
  • Melbourne Research Scholarship: Full fee offset and stipend
  • Second Prize - Electronic Trading Competition - Jane Street Capital (2019)

Activities

  • Presenter at Australasian Joint Conference on Artificial Intelligence (AJCAI) Doctoral Consortium - Perth, Australia - 12/2022
  • Participant at FAccT, Innovative Methods for Critical Studies of Emerging Technologies Workshop - Australian National University, Canberra - 09/2022
  • Final round participant at Melbourne Facebook Hackathon, Australia - 05/2019

The original source code of this website is from here.