How do we make good decisions in the presence of uncertainty? This question arises in numerous contexts, including natural resources management, games, and robot planning & control. The past few decades have seen significant advances in decision-making under uncertainty. These range from new domain-independent methods in areas such as artificial intelligence, statistics, operations research, robot planning, and control theory, to novel domain-specific methods in fields such as ecology, fisheries, economics, and mathematical finance. Unfortunately, progress in one domain may often be easily overlooked by researchers from another community.

This workshop aims to provide a multidisciplinary forum for researchers from disparate fields to discuss recent advances in decision making, identify research challenges, and explore potential collaborations.

We invite contributed talks on topics including but not limited to the following:

Decision models (e.g., Markov Decision Processes (MDPs), POMDPs)

Decision theory (e.g., expected utility theory, bounded rationality)

Planning under uncertainty

Reinforcement learning

Stochastic control (e.g., LQG, robust control)

Operations research

Applications (e.g., natural resource management, robot autonomy, pandemic management, natural disaster response, portfolio management)

We also invite submissions of original research articles to a special issue of the Annals of Operations Research, which is on the same topic as the workshop. A CFP for the special issue will be forthcoming.

Register using the link below to join us in the workshop.

Iadine Chadès, Principal Research Scientist, CSIRO Title: Developing ML and AI Decision Tools for Conservation Abstract:I will give an overview of the ML and AI research we have been conducting to help make better decisions in the field of conservation (adaptive management) over the last 10 years. In particular, we have developed algorithms to solve and increase interpretability of Markov decision models and stochastic dynamic programming. I will be highlighting that the current bottleneck is not necessary our ability to solve complex decision problems, rather it is our ability to make solutions easy to interpret and more likely to be trusted. Bio: Iadine's research is at the forefront of linking domain sciences such as ecology, epidemiology, synthetic biology with quantitative tools from the field of artificial intelligence (AI). She develops AI methods to provide guidance on how to make smart decisions under imperfect knowledge and resource constraints. With CSIRO, Iadine is currently an activity leader with the MLAI Future Science Platform (2019-) where she focuses on developing ML/AI for decision-making. She is a Chief Investigator with NHMRC Centre for Research Excellence SPECTRUM (Supporting Participatory Evidence generation to Control Transmissible diseases in our Region Using Modelling). From 2012 to 2021, she was the team leader of the Conservation Decisions team (CSIRO, Land and Water) a multisdisciplinary group with expertise in ecology, systematic conservation planning, priority threat management, artificial intelligence, and decision theory.

Alan Hájek, Professor, Australian National University Title: Ω Abstract: Probability theory is the dominant approach to modeling uncertainty. We begin with a set of possibilities or outcomes, usually designated ‘Ω’. We then assign probabilities — real numbers between 0 and 1 inclusive — to subsets of Ω. Nearly all of the action in the mathematics and philosophy of probability for over three and a half centuries has concerned the probabilities: their axiomatization, their associated theorems, and their interpretation. My most recent project regarding probability is to put Ω in the spotlight; this is a progress report. Ω is a set of possibilities, but which possibilities? While the probability calculus constrains our numerical assignments, and its interpretation guides us further regarding them, we are entirely left to our own devices regarding Ω. What makes one Ω better than another? Its members are typically not exhaustive — but which possibilities should be excluded? Its members are typically not maximally fine-grained — but how refined should they be? I will discuss both philosophical and practical problems with the construction of a good Ω. Along the way, I will question the notion of a ‘catch-all’ that is supposed to cover all possibilities that have not been explicitly identified. I will end with an omega dilemma: roughly, either orthodox probability theory breaks down, and must be rethought; or we need an account of how probabilities should be revised when a proposition that is not in Ω is learned. Bio: Alan Hájek's research interests include the philosophical foundations of probability and decision theory, epistemology, the philosophy of science, metaphysics, and the philosophy of religion. His paper "What Conditional Probability Could Not Be" won the 2004 American Philosophical Association Article Prize for "the best article published in the previous two years" by a "younger scholar". The Philosopher's Annual selected his "Waging War on Pascal's Wager" as one of the ten best articles in philosophy in 2003. He is a Fellow of the Australian Academy of the Humanities. He was the keynote speaker at the 2007 Chinese Analytic Philosophy Association conference, Wuhan. He was the President of the Australasian Association of Philosophy in 2009-2010. He received the 2012 Award for Excellence in Supervision, ANU College of Arts and Social Sciences. In 2013 he won the ANU-wide Vice-Chancellor's Award for Excellence in Supervision.

The full program can be downloaded here

All times are in Australian Eastern Standard Time

08:30	Registration
09:00	Opening Address: Prof. Jenny Seddon, Acting Executive Dean, FoS, UQ
Keynote
09:05	Alan Hájek
09:50	Coffee
Session A. Ubiquitous Uncertainties
10:20	Dragan Rangelov	Perceptual Decision Making Relies on Reducing Uncertainty about Neural Sensory Representations
10:40	Antonio Rosato	Quality is in the Eye of the Beholder: Taste Projection in Markets with Observational Learning
11:00	Frankie Cho	How Uncertainty Changes Optimal Decisions for National Land Use Change
11:20	Break
Session B. Operations Research
11:30	Michael Forbes	An Exact Algorithm for the Pickup and Delivery Problem with Time Windows and Demand Uncertainty
11:50	Rick Jeuken	Active Set Methods for Solving Large Sample Average Approximations of Chance Constrained Optimisation Problems
12:10	Kazutoshi Yamazaki	On the CUSUM Procedure for Phase-Type Distributions: A Levy Fluctuation Theory Approach
12:30	Lunch
Keynote
13:30	Iadine Chadès
14:15	Coffee
Session C. Planning
14:45	Marcus Hoerger	Adaptive Discretization using Voronoi Trees for Continuous-Action POMDPs
15:05	Luz Pascal	A Universal 2-state n-action Adaptive Management Solver
15:25	Nicholas Collins	Locally-Connected Interrelated Network: A Forward Propagation Primitive
15:45	Break
Session D. Reinforcement Learning
15:55	Jun Ju	Model-based Offline Reinforcement Learning for Sustainable Fishery Management
16:15	Vektor Dewanto	Discounting-Free Reinforcement Learning from Transient States
16:35	Konstantin Avrachenkov	Full gradient DQN Reinforcement Learning: A Provably Convergent Scheme
16:55	Closing Remarks

The workshop will be held in the Boardroom of St Leo's College at the University of Queensland. St Leo's College is located on College Rd, St Lucia QLD 4067, shown on the map below. Remote participants and speakers can join the workshop via Zoom. The Zoom link will be announced shortly.

Buses

Translink provides safe and convenient bus options to get from and to St Leo's college. The bus stops closest to the venue are UQ Lakes Busstation and UQ Chancellor's place.

Parking

Casual parking is available across the campus for 5 AUD per day within the blue parking zones in UQ's parking map.

Wifi

Wifi access is provided via the campus-wide Eduroam network.

Submit contributed talks at https://tinyurl.com/y47ez5a4. Please still register using the registration link below.

Deadline for submission: 17 June 2022
Notification of accepted submissions: 21 June 2022

Please register at https://tinyurl.com/y2erdlwh. If you are attending in-person, please register before 4 July for catering purpose. Registration for in-person attendance may close before 4 July if we reach the maximum capacity.

Nan Ye, UQ
Hanna Kurniawati, ANU
Marcus Hoerger, UQ
Dirk Kroese, UQ
Jerzy Filar, UQ

Please email dmuu2022@gmail.com if you have any question about the workshop.

The University of Queensland
The Australian Research Council