The AAAI organizing committee has decided that all sessions will be held as a Virtual event. The SafeAI workshop schedule has been optimized for allocating suitable time slots to the speakers’ time zones.

Day 1: Scheduled on Feb 28, 2022 from 13:00 to 18:10 UTC [5:00-10:10 PST]

Location: AAAI Virtual Venue, Red Building Room 5.

Time (UTC)


13:00-13:05 Welcome and Introduction – Chair: Gabriel Pedroza (CEA List)

Keynote 1: Matthew Dwyer (University of Virginia), Distribution-aware Test Adequacy for Neural Networks

13:50-14:00 Short Break
Special Session 1: EnnCore – Chair: Lucas Cordeiro (University of Manchester)

EnnCore addresses the fundamental problem of guaranteeing safety, transparency, and robustness in neural-based architectures.

Coffee Break

Invited Talk 1: Shiri Dori-Hacohen (University of Connecticut), Quantifying Misalignment Between Agents

Technical Session 1: Bias, Fairness and Value Alignment – Chair: José Hernández-Orallo (Universitat Politècnica de València)

The Problem of Behaviour and Preference Manipulation in AI Systems, Hal Ashton and Matija Franklin.
IFBiD: Inference-Free Bias Detection, Ignacio Serna, Daniel DeAlcala, Aythami Morales MorenoJulian Fierrez and Javier Ortega-Garcia.
Blackbox Post-Processing for Multiclass Fairness, Preston Putzel and Scott Lee.

– Debate Panel – Paper Discussants: Borja G. Leon (Imperial College London), Richard Mallah (Future of Life Institute)

17:20-17:30 Short Break
Technical Session 2: Interpretability and Accountability – Chair: John McDermid (University of York)

A Gray Box Model for Characterizing Driver Behavior, Soyeon Jung, Ransalu Senanayake and Mykel Kochenderfer.
Defining and Identifying the Legal Culpability of Side Effects using Causal Graphs, Hal Ashton.

– Debate Panel – Paper Discussants: Peter He (University College London), Roel Dobbe (TU Delft)

Poster Pitches 1 – (2 mins x pitch)

Near-Term AI as an Existential Risk Factor, Ben Bucknall and Shiri Dori-Hacohen.
The Dilemma Between Data Transformations and Adversarial Robustness for Time Series Application Systems, Sheila Alemany and Niki Pissinou.
Interpretable Local Tree Surrogate Policies, John Mern, Sidhart Krishnan, Anil Yildiz, Kyle Hatch and Mykel J. Kochenderfer.

18:05-18:10 Wrap-up Day 1
18:10-18:40 Poster Exhibition


Day 2: Scheduled on Mar 1, 2022 from 8:00 to 17:15 UTC [0:00-9:15 PST]

Location: AAAI Virtual Venue, Red Building Room 5.

Time (UTC)



Invited Talk 2: Roel Dobbe (TU Delft), A System Safety Perspective for Developing and Governing Artificial Intelligence

Technical Session 3: Robustness and Uncertainty – Chair: Xin Cynthia Chen (University of Hong Kong)

Efficient Adversarial Sequence Generation for RNN with Symbolic Weighted Finite Automata, Mingjun Ma, Dehui Du, Yuanhao Liu, Yanyun Wang and Yiyang Li.
A Study on Mitigating Hard Boundaries of Decision-Tree-based Uncertainty Estimates for AI Models, Pascal Gerber, Lisa Jöckel and Michael Kläs.
Quantifying the Importance of Latent Features in Neural Networks, Amany Alshareef, Nicolas Berthier, Sven Schewe and Xiaowei Huang.
Maximum Likelihood Uncertainty Estimation: Robustness to Outliers, Deebul Nair, Nico Hochgeschwender and Miguel Olivares-Mendez.

– Debate Panel – Paper Discussants: Xiaowei Huang (University of Liverpool), Mauricio Castillo-Effen (Lockheed Martin)

Poster Pitches 2 – (2 mins x pitch)

Oases of Cooperation: An Empirical Evaluation of Reinforcement Learning in the Iterated Prisoner’s Dilemma, Peter Barnett and John Burden.
Leveraging Multi-task Learning for Umambiguous and Flexible Deep Neural Network Watermarking, Fangqi Li, Lei Yang, Shilin Wang and Alan Wee-Chung Liew.
Human-in-the-loop Learning for Safe Exploration through Anomaly Prediction and Intervention, Prajit T Rajendran, Huascar Espinoza, Agnes Delaborde and Chokri Mraidha.
Safety Aware Reinforcement Learning by Identifying Comprehensible Constraints in Expert Demonstrations, Leopold Müller, Lars Böcking and Michael Färber.

9:45-10:15 Coffee Break & Poster Exhibition
Special Session 2: – Chairs: Bertrand Braunschweig (IRT SystemX), François Terrier (CEA List), Rodolphe Gélin (Renault Group TCR) is the largest initiative in Europe for developing a software platform for trustworthy AI engineering.

Lunch Break

Invited Talk 3: Bonnie W. Johnson (Naval Postgraduate School), Safety in AI-Enabled Warfare Decision Aids

Technical Session 4: Safe Reinforcement Learning – Chair: Richard Mallah (Future of Life Institute)

Reinforcement Learning With Imperfect Safety Constraints, Jin Woo Ro, Gerald Lüttgen and Diedrich Wolter.
Do Androids Dream of Electric Fences? Safety-Aware Reinforcement Learning with Latent Shielding, Peter He, Borja G. Leon and Francesco Belardinelli. [Best Paper Award]
HiSaRL: A Hierarchical Framework for Safe Reinforcement Learning, Zikang Xiong, Ishika Agarwal and Suresh Jagannathan.
A Game-Theoretic Perspective on Risk-Sensitive Reinforcement Learning, Mathieu Godbout, Maxime Heuillet, Sharath Chandra Raparthy, Rupali Bhati and Audrey Durand.

– Debate Panel – Paper Discussants: Ben Bucknall (Uppsala University), John Burden (University of Cambridge), Gabriel Pedroza (CEA List)

Poster Pitches 3 – (2 mins x pitch)

Combining Data-Driven and Knowledge-Based AI Paradigms for Engineering AI-Based Safety-Critical Systems, Juliette Mattioli, Gabriel Pedroza, Souhaiel Khalfaoui and Bertrand Leroy.
Is it all a cluster game? – Exploring Out-of-Distribution Detection based on Clustering in the Embedding Space, Poulami Sinhamahapatra, Rajat Koner, Karsten Roscher and Stephan Günnemann.
A Practical Overview of Safety Concerns and Mitigation Methods for Visual Deep Learning Algorithms, Saeed Bakhshi Germi and Esa Rahtu.
Comparing Vision Transformers and Convolutional Nets for Safety Critical Systems, Michal Filipiuk and Vasu Singh.
A Framework to Argue Quantitative Safety Targets in Assurance Cases for AI/ML Components Combining Design and Runtime Safety Measures, Michael Klaes, Lisa Jöckel, Rasmus Adler and Jan Reich.

14:50-15:20 Coffee Break and Poster Exhibition

Keynote 2: Ganesh Pai (KBR / NASA Ames Research Center), Towards Certification of Machine Learning in Aeronautical Applications

Technical Session 5: AI Testing and Assessment – Chair: Mauricio Castillo-Effen (Lockheed Martin)

Beyond Test Accuracy: The Effects of Model Compression on CNNs, Adrian Schwaiger, Kristian Schwienbacher and Karsten Roscher.
Differential Assessment of Black-Box AI Agents, Rashmeet Kaur Nayyar, Pulkit Verma and Siddharth Srivastava.
Using Adaptive Stress Testing to Identify Paths to Ethical Dilemmas in Autonomous Systems, Ann-Katrin Reuel, Mark Koren, Anthony Corso and Mykel J. Kochenderfer.

– Debate Panel – Paper Discussants: Bonnie Johnson (Naval Postgraduate School), José Hernández-Orallo (Universitat Politècnica de València)


Wrap-up and Best Paper Award