Program

Day 1  (Wednesday, 1.2.23)

9:00 Gathering

9:30 Opening Remarks

10:00 Gal Chechik (BIU)

Talk about images: Learn to reason about the perceived world 

AI aims to build systems that can interact with their environment, with people, and with other agents in the real world. This vision poses hard algorithmic challenges for learning. Such systems are often required to learn to generalize effectively from few samples, to communicate their understanding in ways that are natural to people, and "use common sense" to take into account the unseen context of an image. I will discuss several research thrusts for facing these challenges, from processing scene graphs, through learning by elimination to personalizing generative AI models. 

11:00 Coffee Break

11:15 Session: ML and applications

• Draw Me a Flower: Processing and Grounding Abstraction in Natural Language / Royi Lachmy, Valentina Pyatkin, Avshalom Manevich, Reut Tsarfaty

• QASem Parsing: Text-to-text Modeling of QA-based Semantics / Ayal Klein, Eran Hirsch, Ron Eliav

• Cross-document Event Coreference Search: Task, Dataset and Modeling / Alon Eirew, Avi Caciularu, Ido Dagan

• What drives performance in machine learning models for predicting heart failure outcome? / Rom Gutman, Doron Aronson, Oren Caspi, Uri Shalit

• Bounded Future MS-TCN++ for surgical gesture recognition / Adam Goldbraikh, Netanell Avisdris, Carla M. Pugh, Shlomi Laufer

• Incorporating time-interval sequences in linear TV for next-item prediction / Veronika Bogina, Yuri Variat, Tsvi Kuflik, Eyal Dim

12:45 Lunch

13:30 IAAI Business Meeting

14:00 Yuval Shahar (BGU)

Reducing The Noise in Medical Care Through the Integration of Data Science and Symbolic Reasoning
Modern medical care has made several giant strides within the past century. However, in addition to insufficient knowledge regarding multiple disorders, medical care suffers also from significant noise (unjustifiable inter-provider and intra-provider variance), several types of cognitive biases, and limitations due to human bounded rationality regarding the capability to access and process multivariate, longitudinal patient data, the capability to retrieve the context-sensitive knowledge most relevant to processing it, and the capability to correctly compute the optimal strategy for each patient. These deficiencies exist even in domains in which sufficient knowledge exists regarding best practice guidelines.
At the core of medical care lie several fundamental tasks such as context-sensitive monitoring, interpretation, and analysis of large amounts of time-stamped clinical data, arriving from multiple sources, and the application of appropriate evidence-based procedural knowledge to manage the patient as needed, taking into consideration also their personal characteristics and references.
Examples in which the fundamental tasks are crucial include the management of chronic patients using evidence-based clinical guidelines, the retrospective assessment of the quality of the application of such a guideline, and the learning of new knowledge from analyzing multivariate, time-oriented clinical data, which supports tasks such as clustering, classification, and prediction.
Provision of automated support to the appropriate management of patients requires an integration of a data-driven analysis of large numbers of multivariate longitudinal patient data, with the application of procedural and declarative medical knowledge, while considering also personal patient contexts and preferences.
The talk describes several conceptual and computational architectures developed by my research teams at Stanford and Ben Gurion universities, for performance of these tasks, as well as several rigorous assessments of these architectures in the USA, Europe, and Israel, to demonstrate how the amount of several types of noise inherent to medical care can be reduced.

15:00 Session: HCI and HAI

• Tell me something interesting: Clinical utility of machine learning prediction models in the ICU / Bar Eini-Porat, Ofra Amir, Danny Eytan, Uri Shalit

• Considering temporal aspects in recommender systems: a survey / Veronika Bogina, Tsvi Kuflik, Dietmar Jannach, Maria Bielikova, Michal Kompan, Christoph Trattner

• Enhancing Fairness Perception – Towards Human-Centred AI and Personalized Explanations Understanding the Factors Influencing Laypeople’s Fairness Perceptions of Algorithmic Decisions / Avital Shulner-Tal, Tsvi Kuflik, Doron Kliger

• “I Don’t Think So”: Summarizing Policy Disagreements for Agent Comparison / Yotam Amitai, Ofra Amir

• A (dis-)information theory of revealed and unrevealed preferences / Nitay Alon, Lion Schulz, Jeffrey S. Rosenschein, Peter Dayan

• CAMS: Collision Avoiding Max-Sum for Mobile Sensor Teams  / Arseni Pertzovskiy, Roie Zivan and Noa Agmon 

16:30 Poster Session

Day 2  (Thursday, 2.2.23)

9:00 Gathering

9:30 Awards

10:00 Michal Feldman (TAU)

Algorithmic Contract Design 

Contract design captures situations where a principal delegates the execution of a costly task to an agent. To complete the task, the agent chooses an action from a set of costly actions. The principal can only observe the outcome, which is stochastically determined by the chosen action. The principal incentivizes the desired action through a contract, that specifies payments based on the observed outcome. In this talk, I will survey two papers on *combinatorial* contracts, which highlight different sources of complexity that arise in contract design. The first (FOCS’21) is where the agent can choose any subset of a given set of actions; the second is where the principal motivates a team of agents. We provide (approximation) algorithms and hardness results for the optimal contract problem in these scenarios.

Based on joint work with Tomer Ezra, Paul Duetting and Thomas Kesselheim.

11:00 Coffee Break

11:15 Session: Interaction and Resource Allocation

• Resource Sharing Through Multi-Round Matchings / Yohai Trabelsi, Abhijin Adiga, Sarit Kraus, S. S. Ravi, Daniel J. Rosenkrantz

• Asynchronous Communication Aware Multi-Agent Task Allocation / Ben Rachmut, Sofia Amador Nelke and Roie Zivan

• Explainable Reinforcement Learning via Model Transforms / Mira Finkelstein, Lucy Liu, Nitsan Levy Schlot, Yoav Kolumbus, David C. Parkes, Jeffrey S. Rosenshein, Sarah Keren 

• Fair and Truthful Giveaway Lotteries /  Yonatan Aumann, Tal Arbiv 

• Challenges in the task of unsupervised features selection for high-dimensional data / Dan Vilenchik

12:45 Lunch

14:00 Session: MAS

• Incentive-based Efficient Solutions for Games on Networks   / Yair Vaknin

• Distributed Spectrum-based Fault Localization / Avraham Natan, Roni Stern, Meir Kalech

• Ask and You Shall be Served: Representing and Solving Multi-agent Optimization Problems with Service Requesters and Providers / Maya Lavie, Tehila Caspi, Omer Lev and Roie Zivan 

• Privacy Preserving DCOP Solving by Mediation / Tamir Tassa, Pablo Kogan, Avishay Yanai, Tal Grinshpoun

• Complexity of Probabilistic Inference in Random Dichotomous Hedonic Games / Sa'ar Cohen, Noa Agmon

• Graph Search Theory and Algorithms for Planning with Dynamically Estimated Action Models  / Eyal Weiss, Gal Kaminka

15:30 Poster Session