AAAI-23 Bridge Program
Thirty-Seventh AAAI Conference on Artificial Intelligence
February 7 – 8, 2023
Walter E. Washington Convention Center
Washington DC, USA
Sponsored by the Association for the Advancement of Artificial Intelligence
AAAI-23 Bridge Program
(The Bridge Program schedule will be available in November 2022.)
The community of researchers at the intersection of AI and Finance has been growing steadily, but can benefit from increased exposure, especially at premier AI venues like AAAI. Three years ago, a new conference was established with ACM: The ACM International Conference on AI in Finance (ICAIF). Since that time, ICAIF has been the major forum of discussion and exchange of research ideas in the use of AI in financial applications. Now, it is the right time for the whole AI community to consider addressing the open questions in the use of AI in this domain.
The goal of this bridge is to bring together AI researchers and practitioners from industry and academia, to share technical advances and insights of the application of AI techniques to financial services in the private sector. The target audience is AI researchers that are actively working on the use of AI in financial private institutions as well as researchers that would like to explore the potential application of their work to this domain. From the industry side, we are open to the participation of researchers or professionals that would like to understand the potential application of AI to their business.
Note: This bridge aims to foster collaboration and a productive exchange of ideas and experiences among private entities with an economic role; conversely, if you work for a public financial institution, you will find a suitable forum in the companion AAAI-23 bridge titled “AI and Financial Institutions”, to which you are encouraged to submit.
All different types of submissions accept contents on topics that are relevant to general financial problems that may include but are not limited to:
- Generative models and data-driven simulation
- Planning, Search, Constraint-based Reasoning, Optimization, and Reinforcement learning
- Meta learning, federated learning, representation learning and transfer learning
- Natural language processing
- Time series prediction
- Validation and calibration of financial models
- Multi-agent systems and game-theoretic analysis of financial markets
- Explainability, ethics, and fairness of AI & ML systems
- Security, and privacy of AI & ML systems
- Computational regulation and compliance in finance
- Robustness and uncertainty quantification
- Potential applications of interest may include but are not limited to:
- Fraud detection for credit cards and mortgages
- Early detection of firm defaults
- Blockchain and cryptocurrency
- Risk modelling and risk management
- Trading (e.g., optimal execution, market making, smart order routing and hedging)
- Pricing strategies
- Robot-advising and investment recommendations
- Forecasting of financial scenarios
- Financial time series analysis and factor models
- Chatbots, automated analysis of documents
The bridge will be composed of several types of activities: one/two tutorials, one/two invited talks, paper presentations, panel discussions, student poster presentations, and students mentoring.
The bridge would like to receive different types of submissions:
- Tutorials that introduce the main open problems within AI/financial applications. Send proposals of tutorials to firstname.lastname@example.org. They should include
o Topic of tutorial
o Description of its contents
o Names and shot bios of presenters
o Length (up to two hours)
- Original short papers that showcase the current state of the application on AI in financial services. They should be 4 pages long in AAAI format. We encourage students to submit their current work.
- Papers published in other venues. We require authors to upload an extended abstract of 2 pages specifying in which venue the paper was accepted.
Link to submissions: https://cmt3.research.microsoft.com/AIFinBridge2023/Submission/Index
Updated Submission Deadline: November 23, 2022
- Tucker Balch, Ph.D., Managing Director, J.P. Morgan AI Research, and Adjunct Professor, Georgia Institute of Technology, email@example.com
- Daniel Borrajo, Ph.D., Executive Director, J.P. Morgan AI Research, and Professor at Universidad Carlos III de Madrid, firstname.lastname@example.org
- Julia Stoyanovich, Ph.D., Associate Professor, New York University, email@example.com
- Susan Tibbs, J.D., Vice President, Market Regulation, Financial Industry Regulatory Authority, Susan.Tibbs@finra.org
- Manuela Veloso, Ph.D., Managing Director, Head, J.P. Morgan AI Research, and Herbert A. Simon University Professor, Emerita, Carnegie Mellon University, firstname.lastname@example.org
- Carmine Ventre, PhD., Professor Computer Science, King’s College London, email@example.com
- Larry Wall, Ph.D., Executive Director, Center for Financial Innovation, Federal Reserve Bank of Atlanta, firstname.lastname@example.org
Automated Program and Proof Synthesis (AP2S) are two long-standing, closely related challenges in AI, recently advanced through the incorporation of deep learning. However, much research focuses exclusively on one or the other, instead of leveraging synergies between them. This bridge program aims to connect researchers from each sub-field and educate them about the other, including but not limited to recent deep learning techniques.
The AP2S bridge welcomes contributions in the form of educational lectures/tutorials and extended research abstracts. The majority of the program will consist of 30–45-minute lectures/tutorials, grouped by topic into 90-minute sessions. Time in between sessions will be allocated for poster presentations of accepted abstracts. The final session will include brief 5-minute lightning talks for selected posters, a group panel discussion, and a speed-meeting networking event.
The focus of the bridge is on underlying principles common to both program and proof synthesis. Both problems involve document generation in strict formal languages, leveraging large structured prior “knowledge” (known theorems or existing software libraries), navigating vast search spaces (possible proofs and programs), and high-level cognitive abilities (creativity and abstraction). Both problems have witnessed recent advances with deep learning, making this program especially timely. Each problem can also be framed as an example of the other.
Grand challenges envisioned for this bridge include:
- Automatically synthesized constructive existence proofs for novel algorithms with improved complexity characteristics
- Deciding as-yet unsolved conjectures by automatically synthesizing programs in the languages of interactive proof systems
- “Executing” a partially developed proof on concrete examples as “input”, similarly, to running a program, to guide the automated proof synthesis process.
- State-of-the-art in both program and proof synthesis using a common core deep learning technique
- Fast, automated program and proof repair when a dependency (software library or axiomatization) is changed
Topics for educational lectures/tutorials include:
- Automated theorem proving techniques (e.g., resolution, paramodulation, term rewriting, etc.)
- Automated program synthesis methods (e.g., abstraction domains, genetic programming)
- Interactive proof systems (e.g., Metamath, Lean, Coq, Isabelle/HOL, etc.)
- Automated proof systems (e.g., Vampire, E, etc.)
- Deep learning techniques relevant to AP2S (e.g., transformers, graph neural networks, etc.)
- Recent datasets relevant to AP2S (e.g., HOLStep, ProgRES)
- Human reasoning processes during human program and/or proof synthesis
Extended abstracts may present current research or positions on:
- Program and/or proof synthesis, automated and/or in humans, and with or without deep learning.
- New datasets and benchmarks for AP2S
- New directions or practices important to advance AP2S going forward
The bridge will be a one-day program with the following format:
- Quarter-day session with lectures/tutorials related to automated theorem proving
- Quarter-day session with lectures/tutorials related to automated program synthesis
- Quarter-day session with lectures/tutorials related to relevant deep learning techniques
- 5-minute lightning talks for selected posters
- Group panel discussion on AP2S going forward
- Speed-meeting networking session and social hour
Breaks in between sessions will be used for poster presentations of accepted abstracts.
The bridge aims to include 6-9 educational lectures/tutorials, 4-6 panel speakers, and 10-20 accepted posters, 5 of which will be selected for lightning talks. Space permitting, all are welcome to attend, with priority given to speakers, authors of accepted abstracts, their collaborators, and students.
Submissions are accepted in either of the following categories:
- Proposals for educational lecture/tutorials should include draft presentation slides in PDF format, CVs of the speaker(s) in 2-page NSF Biosketch Format, and a brief cover page. The cover page must include a paragraph summary of the talk, a proposed duration (either 30, 45, or 90 minutes), and the names, affiliations, and contact information for the speaker(s).
- Extended abstracts should conform to the AAAI paper guidelines and be a maximum of 4 pages excluding references. Abstracts will be accepted for poster presentation and a subset will be selected for 5-minute lightning talks at the bridge meeting.
Submissions should be made on EasyChair at this link:
- Garrett Katz, Syracuse University, email@example.com
- Kristopher Micinski, Syracuse University, firstname.lastname@example.org
For a copy of this CFP and more information (confirmed speakers, program schedule, recorded talks, etc.) see this web page for the bridge:
The fields of causality and continual learning focus on complementary aspects of human cognition and are fundamental components of artificial intelligence if it is to reason and generalize in complex environments. On the one hand, causality theory provides the language, algorithms, and tools to discover and infer cause-and-effect relationships from data and a partial understanding of a complex system. On the other hand, continual learning systems learn over time from a continuous stream of data, enable knowledge transfer, and alleviate potentially malicious interference when distributional shifts are experienced. However, despite some recent interest in bringing the two fields together, it is currently unclear how causal models may describe continuous streams of data and, vice versa, for continual learning to exploit learned causal structure.
We propose to bring together the fields of continual learning and causality in a two-day AAAI-23 bridge program called “Continual Causality”. In this bridge program, we aim to take the first steps towards a unified treatment of these fields and to provide the space for learning, discussions, and to connect and build a diverse long-term community. To this end, the Continual Causality Bridge invites submissions that present general positions and visions of how to link the two fields, outline challenges that need to be overcome, highlight synergies, and propose tangible future steps. Submissions of position papers should be up to two pages (excluding references, and without appendices) in the AAAI format. The submission deadline is November 18th, 2022 (AOE).
To provide some suggestions, position papers could center around the following aspects, including but not limited to:
- Continual learning and exploitation of causal systems in dynamic non-stationary environments.
- Catastrophic interference and knowledge transfer in learning causal models in the context of continuous streams of data.
- Effective ways for causal structure to aid in leveraging the accumulated knowledge of a continual learning system.
- Leveraging causal tools to interpret distributional shifts in continual learning.
- Next generation benchmarks that go beyond repurposing of existing datasets to support the above items and further essential research questions towards a symbiosis of continual learning and causality.
As a two-day event (see detailed timeline in the program tab), our bridge offers a varied set of activities that support education and discussion at various levels of expertise. Educational activities involve two traditional tutorials, as an entry to the respective fields of continual learning and causality, for both newcomers and experts that are only familiar with either of them. Similarly, two software labs will provide respective hands-on experience on the practical level, derived off the popular emerging software tools Avalanche and DoWhy.
Following these educational activities, a set of vision talks on each day will provide positions on how the fields of continual learning and causality can be brought further together, why a necessity for the latter exists, and what respective challenges may need to be faced in the imminent future. The first of these sets intends to leverage the expertise of researchers with an increased level of seniority, who will exploit their broad expertise towards first attempts at a more integrated vision. The second set will include contributed talks, based on the bridge’s call for contributions of two-page position papers by the community, which will be accepted on the grounds of a peer-review process. Apart from a general poster session, a select set of papers with exceptional level of clarity will be invited to share their views.
Finally, the invited vision talks and contributed position works will serve as a basis for interactive breakout sessions towards the end of the bridge. Participants of the bridge will get the chance to network, discuss presented ideas, and delve into a deeper exchange in more detailed conversation. The various outcomes of this session will ultimately be collected and disseminated beyond the bridge attendees.
Our Continual Causality Bridge aims at bringing together researchers, students, and practitioners interested in causality and continual learning, providing a unique opportunity to discuss ideas, challenges, resources, and opportunities in bridging these two fields. Our initial ideas of research items are intended to facilitate discussion among participants about how continual learning and causality may link together. Ideas are intentionally not yet exhaustive and are open for interpretation by the community. The various proposed interactive bridge elements and discussion activities are planned as a catalyst to emerge with a set of even more concise ideas to lay out the future of this exciting field as a direct take-away for the audience. The primary goal is thus to target a broad audience and build a long- lasting inclusive community.
All submissions will be managed through OpenReview and will later be collected in a proceedings volume. The review process is double-blind, so submissions should be anonymized. As our bridge’s focus is on building a long-term community, the review process will be light and inclusive, focusing primarily on technical correctness as a means of evaluation. Our vision is for prospective community members to voice diverse views that have the potential to advance AI through an ongoing cross-disciplinary exchange. We therefore have no strict constraints on the exact sub-topics of submissions, as long as they target the overall goal of bridging the fields. Submissions are free to focus on a single particularly interesting synergy, take an angle from a specific scientific discipline, or sketch a grander view.
- Keiland Cooper; University of California; email@example.com
- Devendra Dhami; TU Darmstadt & hessian.AI; firstname.lastname@example.org
- Martin Mundt; TU Darmstadt & hessian.AI; email@example.com
- Tyler Hayes; RIT; firstname.lastname@example.org
- Alexis Bellot; DeepMind; email@example.com
- Adele Ribeiro; Philipps-Universität Marburg; firstname.lastname@example.org
- James Smith; Georgia Tech; email@example.com
Artificial Intelligence (AI) and Robotics have been strongly connected areas since the early days of AI. On the one hand, AI principles and methods play a crucial role in several areas of Robotics research and are pervasively exploited at various levels of robot architectures. On the other hand, Robotics provides several relevant challenges for the AI community where different AI methods and tools require to be deeply integrated to obtain an embodied agent capable of an autonomous, adaptive, and interactive behavior. In this direction, the scientific relevance of this proposal is related to how diverse areas of AI and Robotics research can be integrated to enable robots to face complex challenges in real-world tasks.
Notwithstanding the shared challenges and the widespread learning and statistical methods in robotics, the AI and Robotics communities still need to further interact to integrate methods that are still not fully exploited. In particular, the integration of symbolic and sub-symbolic models and methods with various levels of abstraction are needed to face many relevant issues (e.g., combined task and motion planning, neuro-symbolic learning, task teaching, collaborative task execution, etc.) that require a strict collaboration between the two communities.
Some initiatives have already been started to create a stable, long-term forum where researchers from both AI and Robotics communities can openly discuss relevant issues such as research and development progress, future directions and open challenges related to AI methods in Robotics. A notable venue is the International Conference on Automated Planning and Scheduling (ICAPS) workshop series on Planning and Robotics (PlanRob). Started during ICAPS 2013, in Rome (Italy), and followed by editions at every ICAPS from 2014 to 2021, the PlanRob Workshop series has gathered excellent feedback from the P&S community, as also confirmed by the organisation of a Special Track on Robotics starting with ICAPS 2014. On the Robotics side, a version of the PlanRob series was held at ICRA 2017 in Singapore (top conference for the IEEE Robotics and Automation Society). Some related events were also organized at the European Robotic Forum (promoted by the European Commission) to present results and challenges in EU-funded projects.
After such time, technological developments allow embedded devices to run complex algorithms and be more effective, a wide diffusion of data is enabling opportunities for ground-breaking impact from learning techniques when applied to robots and an increasing number of AI & Robotics results are being collected facilitating the deployment of intelligent robots in many applicative fields.
The goal of this Bridge Session is to keep fostering the interactions already ongoing within PlanRob and further stabilize such common ground where researchers can confront and broaden research frontiers, identify most crucial challenges, and pave the way towards a shared research agenda. Gathering notable representatives from both communities in a unique venue contributes to achieving such an objective and allows them to share their knowledge and expertise as well as take advantage of different perspectives.
We look for position papers to discuss ideas and concepts, e.g., via critiques or new perspectives in AI and Robotics, historical perspectives and analysis. Some relevant areas are, e.g., long-term autonomy and robot learning, neural symbolic approach, task teaching, cognitive robotics, human-robot interaction, open world planning and acting, explainable AI in robotics, reliable and safe AI methods for robotics, technological and integration challenges. The list is no limite to the above areas and additional topics may be proposed by contributors.
We are also looking for tutorials, a survey presentation, or an integrated application showcase proposals on (new) cross-disciplinary AI & Robotics topics, with a shared focus, and to engage and educate new researchers, as well as established researchers. The goal is also to set common grand challenges while discussing fundamental problems and educating each other about respective tools.
This session will be a one-day long event and it will be structured as a sequence discussion sessions, tutorials, and interactive discussions to confront ideas and synthesize common views. A poster session will be organized to allow the presentation of open contributions and further stimulate the general discussion. A final panel session will aim to select the most crucial elements and identify a possible common research agenda for future steps.
Both position papers and proposals for tutorial, survey presentation or application showcase can be (max) 4 pages in AAAI style (plus up to one page of references). https://www.aaai.org/Publications/Templates/AnonymousSubmission23.zip
Arthur Bit-Monnot, LAAS-CNRS, France Arthur firstname.lastname@example.org
Alberto Finzi, Naples University “Federico II,” Italy email@example.com
Andrea Orlandini, National Research Council of Italy, Italy firstname.lastname@example.org
Business process management (BPM) comprises a spectrum of modeling and management approaches and tools, including robotic process automation (RPA) workflow, case, and decision management. Recent advancement in Artificial Intelligence (AI) and the sudden outbreak of the COVID pandemic has significantly accelerated the need for companies to adopt digitization and automation. According to a recent McKinsey survey, companies have pushed the time frame for digitizing many aspects of their business by three to four years. As AI techniques mature and become deployed with the fidelity and robustness required for enterprise applications, companies are increasingly looking to consume them as part of their business process and automation tools.
While there have been grass root efforts in both AI and BPM communities to explore topics in relation to AI-infused business processes, there has not been a major effort to bridge these two communities together. The AI4BPM Bridge at AAAI 2023 hopes to bring together academics and industry professionals working at the intersection of artificial intelligence and business process management under the same roof. This two-day event will include invited talks, poster sessions, student outreach, meet and mingle opportunities, hands-on system demonstrations, tutorials, and much more!
The Bridge will focus on the interaction between AI approaches, especially agent-based, planning, and machine learning approaches, and BPM research areas and techniques, especially business process modeling, optimization, automation, and process mining. Technical topics include, but are not limited to:
- Cognitive approaches to Business Process Management
- Agent-based modeling and simulation for BPM
- Knowledge Representation (KR) and reasoning about process specifications
- AI enablement for declarative and hybrid models
- AI-driven modeling and optimization of processes
- Process Mining augmented with AI techniques
- Applications of automated planning techniques for BPM
- Non-traditional AI models and approaches to BPM
- Explainable AI and trustworthy AI for operational support in Process Mining and BPM
- Machine Learning to support workflow management
- Recommender Systems for business processes
- Constraint-based reasoning
- Goal and ontology-driven approaches to process management
- Conversational systems, natural language processing, and human-machine interaction for business process management
- AI techniques for process discovery, conformance checking, prescriptive and predictive monitoring
- AI techniques for clustering and classification of process execution traces
- Machine Learning for event recognition on semi-structured and unstructured data
- Association rule mining, specification mining, and decision mining from process execution traces
- Uncertainty in AI for process management
- Multi-agent systems, strategic reasoning, game theory, and mechanism design for multi-party processes
- Multi-objective optimization, decision-making, and continuous improvement
- AI-based robotic process automation (RPA)
- AI-based enrichment of IoT-enabled processes
- Applications of AI for Blockchain-hosted processes
- Applications of AI in industry-specific processes (e.g., retail, e-commerce, finance, manufacturing, healthcare)
- Social, economic, and business impacts of infusing AI into business processes
This bridge organizing committee solicits three types of contributions from both academia and industry on how AI can be applied in BPM contexts as well as how BPM aspects can have an impact on adapting and tuning AI techniques to temporal and business dimensions of BPM.
- Contributed posters: Participants are encouraged to submit 2-page abstracts on their work to participate in an extended poster and meet-and-greet session. This can be about recently published or ongoing work. If applicable, such submissions must indicate clearly when and where the corresponding papers have been or will be published.
- System demonstrations: The poster session will also feature live system demonstrations of tools and software that are useful to both the AI and BPM communities. Submissions of this type are also required a 2-page abstract but must provide links to GitHub (or equivalent), and a description of resources on how to access and use the tool. Treat this as a demonstration submission to a conference, but specifically on the AI x BPM topic.
- Student contributions: Students working at the intersection of AI and BPM are encouraged to submit 2-page abstracts summarizing their work (either in progress or completed). Students will be given an opportunity to present their work as posters and will also be paired with mentors for dedicated mentoring sessions. Treat this as a doctoral consortium submission to a conference, but specifically on the AI x BPM topic.
The type of submission must be clearly indicated in the abstract. All submissions should be formatted in the AAAI-23 style.
Link to submission: https://easychair.org/conferences/submission_new?a=29756028
Link to style files: https://www.aaai.org/Publications/Templates/AnonymousSubmission23.zip
- Tathagata Chakraborti; email@example.com
- Yara Rizk; firstname.lastname@example.org
- Vatche Isahagian; email@example.com
- Andrea Marrella; firstname.lastname@example.org
- Chiara Di Francescomarino; email@example.com
- Jungkoo Kang; Jungkoo.Kang@ibm.com
Please subscribe to our mailing list at https://ai4bpm.com to receive updated news and announcements.
Bringing together CP (Constraint Programming) and ML (Machine Learning) is an important aspect of the larger goal of integrating Reasoning and Learning. Participants are not expected to have prior experience in both fields, but to have familiarity with each at least at the level of an introductory AI course.
The Bridge is designed to provide attendees with a better, broader sense of where we are and where we should be going, and an opportunity to brainstorm, discuss, debate, and find collaborators. The focus will be on bringing together the traditional AI fields of constraint-based reasoning and machine learning, but participants from related fields of reasoning, optimization, and learning, e.g., SAT, operations research, data mining, will be welcome. It is hoped that this one-day Bridge event will help establish an ongoing community, leading to workshops, special issues, etc.
Submissions are sought for presentations addressing the following themes:
- Where Are We Now?
o Mini-tutorials or mini-surveys. They could be broad or narrow in scope, restricted to a particular subtopic, covering recent work from a particular conference or workshop series, or even introducing or surveying a body of the presenters’ work.
- Where Do We Go from Here?
o Position papers, grand challenges, proposals, roadmaps, questions, needs, objectives, opportunities, obstacles, testbeds, applications, metrics. Proposals for discussion topics, panels, or debates. Introductions to ongoing research projects.
o Presentations by those seeking collaborators or mentors for current or proposed work.
o Mini-surveys of available CP or ML tools, or mini-tutorials on individual tools. Examples of the use of CP tools for ML or vice versa. Also welcome are proposals for “hands on” sessions where participants can be guided through a real-time introduction to a tool they access on their laptops.
- Community Building
o Proposals for special issues, conference tracks, workshops, competitions, tools, libraries, benchmark problems, websites, tutorials, surveys, talks, funding proposals, collaborations, blogs, groups, videos, community outreach, social media, instructional material, syllabi, ontologies, books, apps, testbeds, bibliographies, etc.
- Posters are welcome describing previously published, new, ongoing, or proposed work linking CP and ML.
Given the condensed time frame, submissions need only be one- or two-page abstracts, outlining the proposed presentation, the minimum and maximum time it could occupy, and citing relevant experience/publications of the author(s).
Proposals can be for presentations, or portions of presentations, that have been made elsewhere (e.g. tutorials or surveys), suitably updated and scaled for this Bridge event. Presentations (aside from Tools) need to combine CP and ML in some fashion. All presentations should keep in mind that audience members are not expected to have expertise in both areas. Presentations should “reach across the aisle”, and welcome questions and discussion.
Submissions should be PDFs (of the abstracts or the posters). They should contain the affiliations and contact information of the authors.
Submit via EasyChair: https://easychair.org/conferences/?conf=cpml2023
The development of new materials and production processes and the customization of existing ones is increasingly driven by AI, in particular Bayesian optimization and surrogate modeling. In many cases, materials science has relied on compute-intensive simulations to evaluate the properties of proposed designs, or the effect a change might have. Such simulations do not scale to the vast design spaces that materials scientists explore. Machine learning provides an alternative: properties are approximated through the predictions of surrogate models rather than computed by simulations, orders of magnitude faster.
Both AI and materials science are working on conceptually similar problems — how to efficiently identify the best design choices, be that for a machine learning pipeline or a new material. Yet, there is little collaboration between the communities. The purpose of this Bridge is to bring the communities closer together, facilitate cross-disciplinary collaborations, identify common problems, and develop plans for tackling them. We solicit poster submissions that present novel applications, novel algorithms, or pose challenges at the intersection of AI and materials science, in the widest sense.
Whether it’s a mature system or only an idea, we welcome your submissions.
Areas of interest include, but are not limited to:
- Bayesian optimization
- Reinforcement learning
- surrogate modeling
- neural network approaches and their applications to design new materials and production processes
- optimize existing materials and production processes
- characterize or test materials and monitor the performance of materials.
Please feel free to contact the organizers informally for any questions. Posters will undergo a light review by the organizers for suitability for the Bridge.
Please submit a PDF version of your poster on Easychair at:
- Roman Garnett, Washington University at St Louis, firstname.lastname@example.org
- Patrick Johnson, University of Wyoming, email@example.com
- Jessica Koehne, NASA, firstname.lastname@example.org
- Lars Kotthoff, University of Wyoming, email@example.com
For more information, see the bridge website at https://sites.google.com/view/aimat23/home
The aim of this AAAI bridge – organized by the Bank of Italy – is to connect two communities: the AI community and the “Financial Institution Sector” (FIN ST ). The FIN ST community is populated by authorities, agencies, institutions, and other public entities that play a key role in the economic and/or financial life of their home country. A prime example are central banks, but they are by no means the end of the story: There are Financial/Banking Ombudsmans, Supervision Authorities, Financial Intelligence Units, Privacy Protection Authorities, National Statistics Institutes, and more. These players of the public sector are in the middle of a surge of interest in AI, and can fruitfully exchange views and experiences via this bridge. They all work on economic/financial issues in which certain peculiar search / reasoning / induction / deduction / optimization / planning problems are becoming of paramount importance, with a growing fit among AI technologies, tools, techniques, and business. Usually, AI innovates FIN ST processes, but sometimes FIN ST pushes the boundaries of AI. We expect that those who traverse the bridge from FINST to AI will discover a powerful arsenal of new and possibly unfamiliar AI techniques, described and discussed by leading experts in the field: They will likely be surprised by the reach and variety of what state-of-the-art AI has to offer, beyond the established ML approaches that have seen large adoption as of late. Those who cross the bridge the other way around will get exposed to a surprisingly rich domain, with a lot of opportunities for the application of a wide range of AI techniques on both existing and developing use-cases.
We are open to the participation – as listeners or presenters – of any representatives from FIN ST entities, especially those who deal with economics/finance problems and/or have used or plan to use advanced AI tools and techniques to tackle such problems.
From the AI side: Any researcher, practitioners, or AI expert (or, in fact, any participant to the main AAAI conference) is welcome to join, especially those with an interest in economics finance.
Note: This bridge aims to foster collaboration and a productive exchange of ideas and experiences among public entities with an economic role; conversely, if you work for an investment or retail bank, an insurance company, a financial intermediary, or any other private market player, you will find a suitable forum in the companion AAAI-23 bridge titled “AI for Financial Services”, to which you are encouraged to submit.
We would be delighted to hear from the FIN ST community about projects, experiences, challenges, and prototypes in one or more of the following areas (non-exhaustive list, in alphabetical order):
- AI for cybersecurity and vulnerability analysis
- Blockchain analysis and intelligence
- Chatbots and Virtual Agents
- Data modeling and data wrangling
- Dealing with uncertain and/or incomplete information
- Data quality issues
- Data complexity of reasoning tasks
- Deep fake detection, banknote authentication
- Deep learning and reinforcement learning
- Explainability and Ethics
- Fact checking
- Graph analytics
- Imagery analysis for economics and finance
- Intelligent Robotic Process Automation
- Knowledge Representation and Reasoning
- Knowledge Graphs
- Localization/specialization of AI models to non-English and/or domain-specific natural languages
- Multi-Agent systems simulation and calibration
- Neurosymbolic reasoning
- NLP and ML on hybrid (structured + unstructured) data
- Pre-trained ML models and transfer learning
- Privacy-preserving computations
- Reactive automated reasoning
- Sentiment Analysis
- Smart anonymization or pseudonymization of datasets
- Summarization, entity recognition
- Synthetic dataset and time series generation
- Temporal reasoning
- Topic modeling
Just to offer a few concrete examples, consider the following three use cases.
- Any CB is in charge of defining the monetary policy of its home country or monetary area (i.e., setting the official interest rate and controlling the money supply), and to do so there are certain non-trivial “open market” operations to execute. In particular, it must be ensured that financial entities do not submit assets as collateral which are ultimately guaranteed by issuers which are “too closely linked” with the counterparty itself. In doing these checks, one faces non-tractable search and reasoning problems under uncertain and incomplete information.
- Any Supervision Authority is responsible for (macroprudential) supervision, i.e., for regulating and supervising private banks and financial institutions to guarantee the financial (and price) stability within the reference monetary area. Here, the increasing interconnectedness of the financial world makes it so that the knowledge to be exploited to establish who controls whom (and through which ownership chains and which non-financial or implied links) is nowadays to be delegated to a neurosymbolic AI (mixing inductive and deductive approaches).
- Financial Intelligence Units are often responsible for tracing illegal financial activities related to money laundering and terrorism financing, among others. In this case, a bulk of unstructured, extremely sensitive, and oftentimes classified information has to be learned from by some complex AI; the challenge is that the human intelligence controlling the AI does not have access, for security reasons, to all of the knowledge the AI itself can see and has to reason about.
This bridge is a one-day event held in the context of AAAI-23. It will host 3 sessions in the course of the day. The 1 st session will be a traditional keynote session with invited speakers (TBA); the 2nd and 3rd sessions are open to contributions (see “Contributions and deadlines”).
- Session 1: Invited talks [2 hours, including Q&A – chair Marco Benedetti] Plenary talks will be given by three renowned scientists in the field of AI who have been working in close contact with CBs and/or public economic/financial institutions. They will try to convey what it is like to work with public authorities/institutions in the “economics finance” domain, and what are the typical AI problems (and challenges) one encounters.
- Session 2: Services & Software Demos [2 hours, including Q&A – chair Luigi Bellomarini] This session will showcase 5-10 diverse, interdisciplinary projects at the intersection of AI and FIN ST which are already implemented or in an advanced stage of development (i.e., at least in the PoC stage, with a demo-able implementation). After a brief (15 minutes) plenary overview of all the projects, the actual interaction among attendees and proponents will be live, informal, and person-to-person. Live, hands-on demos are thus expected of all entries in Session 2.
- Session 3: Challenges and Use Cases [2 hours, including Q&A – chair Michela Iezzi] Extended abstracts or papers by FIN ST representatives and/or people in the AI community will be presented, in a shared, poster-like, interactive fashion. Abundant time will be allotted to Q&A.
Submissions will be in PDF and will use the standard AAAI 2-column template; they are to be sent via email to firstname.lastname@example.org by November 18. All submissions will be acknowledged. Please use “AAAI23 Bridge <ORGANIZATION_NAME>” as subject. Submissions will be blind reviewed by at least 3 members of the PC. Original, unpublished research material is welcome, but also results already published elsewhere are fully eligible for this bridge.
Anyone from the FIN ST domain – or any AI researcher who has worked on or has experience in FIN ST themes – may submit one of two types of contributions:
- Session 2: A working prototype of (or a fully industrialized version of) an AI-based system that solves problems in the FIN ST domain. The system will be showcased live during Session 2 of the event. A short (1-page or 2-page) PDF description of the solution/tool will accompany each entry. If the system has been successfully brought to production, a discussion of all the technology transfer and industrialization issues encountered will be appreciated.
- Session 3: An extended abstract (max 2 pages) or a paper (max 6 pages) in the standard AAAI template) concerning a use case, business challenge, or AI project in-the-making (not yet suitable for session 2). These contributions should present AI-intensive solutions to problems that proved to be tough or impossible to solve with traditional IT technologies. In addition, pure “challenges” can be submitted: complex problems arising in the FIN ST domain whose exact or approximate solution is out of reach for non-AI technologies and/or still do be designed.
Problems so complex or involved as to require an advancement in the state of the art in some AI disciplines will be favored in the selection. The relevance for the AAAI community, the difficulty of the problem, and the clarity of exposition will be the major criteria adopted by the PC to accept candidates.
- Marco Benedetti (Bank of Italy) – email@example.com
- Luigi Bellomarini (Bank of Italy) – firstname.lastname@example.org
- Michela Iezzi (Bank of Italy) – email@example.com
- Alessandro Maggi (Bank of Italy) – firstname.lastname@example.org
PC members – Academic
- Alberto Finzi (University of Naples, IT)
- Andrea Calì (Birkbeck University, UK)
- Domenico Lembo (La Sapienza University, IT)
- Danilo Antonino Giannone (Alan Turing Institute, UK)
- Emanuel Sallinger (TU Wien, AT)
- Fabrizio Lillo (Scuola Normale Superiore, IT)
- Fosca Giannotti (Scuola Normale Superiore, IT)
- Francesca Medda (UCL, UK)
- Georg Gottlob (Oxford, UK)
- Giuseppe De Giacomo (Oxford, UK)
- Maurizio Lenzerini (La Sapienza University, IT)
- Piero Poccianti (Italian Association for Artificial Intelligence, IT)
- Paolo Papotti (EURECOM, FR)
- Sahar Vahdati (InfAI, DE)
- Stefano Bistarelli (University of Perugia, IT)
PC members – Institutional
(Affiliation is “Bank of Italy, ICT Directorate”, unless noted otherwise)
- Aldo Glielmo
- Antonino Virgillito (National Revenue Agency, IT)
- Claudia Biancotti (Directorate General for Economics, Statistics and Research; Bank of Italy)
- Davide Magnanimi
- Eleonora Laurenza (Financial Intelligence Unit; Bank of Italy)
- Emanuela Girardi (popAI, IT)
- Erik Teunissen (De Nederlandsche Bank)
- Fabiana Rossi
- Juri Marcucci (Directorate General for Economics, Statistics and Research; Bank of Italy)
- Livia Blasi
- Nuno Pereira (Banco de Portugal)
- Oliver Giudice
- Pietro Franchini (IVASS, the Institute for the Supervision of Insurance, IT)
- Valerio Colitta (Directorate General for Financial Supervision and Regulation; Bank of Italy)
The increased availability of data offers the potential to transform healthcare, using advances in machine learning (ML). However, the application of ML to medicine is fraught with pitfalls. Medical datasets often have errors and confounders. Predictive algorithms can fail to recognize data problems and then can make mistakes – often where human experts would not. Many algorithms fail to account for the context in which data is generated and then make poor recommendations. Additionally, these algorithms can be data inefficient, often requiring hundreds of thousands of examples to train. These issues are barriers to the deployment of ML models in clinical settings. The goal of this bridge program is to enable communication between ML practitioners and clinicians who can deploy ML models to assist medical decisions.
- Strategies for reducing the effect of biases and confounding variables in medicine
- Challenges of model deployment in clinical settings
- Interpretability of predictive models in medicine
- Data-efficient algorithms
- Bias mitigation strategies
- Data quality in healthcare data
Format (One day)
- Two Tutorials
- Two Keynote speakers
- Five ten-minute lightning talks followed by five minutes QA sessions.
- A poster session.
- A panel discussion: Interactive Q&A session with a panel of leading researchers
We welcome submissions of abstracts from ML practitioners, clinical researchers, and clinicians in pdf format (2 pages). They must include purpose, method, results, and discussion/conclusions. A maximum of two figures can be submitted with the abstract. All submissions will be peer-reviewed. Some will be selected for lightning talks and poster sessions.
Url for submissions: https://forms.gle/HutXkPLAk3iqGCme8
Gilmer Valdes, PhD, DABR Associate Professor. UCSF, San Francisco, USA.
- Gilmer Valdes, PhD, DABR Associate Professor. UCSF, San Francisco, USA.
- Yannet Interian, PhD. Associate Professor at USF, San Francisco, USA. <email@example.com>
- Clifton David Fuller, MD, PhD Professor, Division of Radiation Oncology, The University of Texas MD Anderson Cancer Center, Houston, TX <CDFuller@mdanderson.org>
- Elena Doering PhD Student. German Center for Neurodegenerative Diseases, Germany <firstname.lastname@example.org>
The main objective of this Bridge Project is to improve the healthcare services of nurses to their patients by building stronger bridges of communication and collaboration between the communities of AI Technology and nursing, especially during the design of AI applications that are intended to augment work performed by nurses. As an initial part of the AAAI-23 conference, we shall bring together representatives from the communities of AI technology and nursing for one full day of educational human interaction.
Human interaction of nurses with their patients affects, of course, the quality of care provided. However, if human services of nurses are to be augmented via introduction of AI tools, then careful attention during the designing of those tools must be given as well to interaction between nurses and the AI tools. Satisfying this requirement can involve specific topics such as a nurse’s trust of the tool, how quickly the nurse needs the tool to respond, or how well nurse and tool (e.g., in the case of deep learning applications) can communicate.
Our educational one-day event will conduct morning open forum tutorial sessions for review (allowing answers for possible outstanding questions) of prepared material that has been furnished in advance to all participants concerning information that the communities of AI technology and nursing need to share in order to communicate effectively. Afternoon sessions then will focus upon moderated discussion, collaborating to formulate specific Grand Challenges regarding future AI tools appropriate for the participating different types of nursing, followed by summing up of what has been learned and is be reported to the main AAAI-23 conference.
This day of communication and collaboration will furnish the foundation for ongoing post-conference learning activity, assisted by use of our website (https://shapingsmarttechnology.org).
We identify three categories of participating target audiences for this project:
- “Prepared” participants (totaling 40-60 people) will represent AI technology and nursing communities and are understood to be organizations able and willing to prepare, in advance of AAAI-23, tutorial materials satisfying submission requirements specified by our Organizing Committee [Dr. Ted Metzler, OCU; Dr. Lundy Lewis, SNHU; Dr. Elizabeth Diener, OCU; Dr. Susan Barnes, UCO; the Rev. Linda C. Pope, United Methodist Church].
- “Registrant” participants will be individuals interested in our subject who may wish initially just to register with AAAI for attending the one-day event. They also will be able, of course, to participate in its proceedings.
- “Ongoing” participants will be either individuals or organizations first engaging or continuing to engage our ongoing program after AAAI-23, via our website or through outreach from our Organizing Committee or other participants in the program. We are expecting that members of this category progressively will form the largest community within our bridging project.
If you believe that your organization may wish to participate in the Prepared manner, please contact our Organizing Committee Chair, Dr. Ted Metzler at email@example.com for a Preparation Information document which will furnish details of the materials to be submitted for our Organizing Committee’s approval and invitation to participate. Although Registrant participants will not be obligated to complete any preparatory work that is to be approved by our Organizing Committee, they also may obtain useful information at any time by contacting firstname.lastname@example.org.
Ongoing participants will be welcome to contact email@example.com after the AAAI-23 conference for information concerning our Bridging AI Technology and Nursing initiative.
- Dr. Ted Metzler, OCU; firstname.lastname@example.org
- Dr. Lundy Lewis, SNHU
- Dr. Elizabeth Diener, OCU
- Dr. Susan Barnes, UCO
- Rev. Linda C. Pope, United Methodist Church
Our website also will supply timely updates regarding the project’s activities. Although Bridging AI Technology and Nursing has originated in the state of Oklahoma, USA, we certainly welcome international participation and may later label the program BAITAN for ease of global reference.
The rapid development of AI technologies and their increasing ubiquity in society are transforming many aspects of our day-to-day lives and revolutionizing a range of domains such as transportation and healthcare. The widespread use of AI systems in society, however, has given rise to serious concerns about accountability, fairness, privacy, safety, and transparency.
Among the AI research community, there is a growing body of literature devoted to tackling these issues. Meanwhile, legal experts and policymakers are initiating greater efforts to regulate AI technologies. For example, the European Commission published a draft of the Artificial Intelligence Act in 2021. We believe it is imperative to bring together experts in AI and Law to discuss open questions and regulatory challenges, as well as to develop an interdisciplinary research agenda.
This bridge meeting targets two groups of audiences: (i) AI researchers who are interested in engaging with legal perspectives on building AI systems with enhanced accountability, fairness, privacy, safety, and transparency, and (ii) legal scholars who are interested in engaging with AI research perspectives on proposing or reforming legal and regulatory governance models for emerging AI technologies. We envision that the bridge meeting will facilitate interdisciplinary dialogues, connect AI and legal researchers for potential collaborations, and lay a solid foundation for community building.
The bridge meeting will focus on the following topics that have been drawing enormous attention and debate in both the AI and Law communities.
- Accountability and Safety: e.g. What are the mechanisms through which accountability of AI systems can be achieved? How should regulators assure safety and efficacy of safety-critical AI systems such as medical devices or autonomous vehicles?
- Explainability and Transparency: e.g. What constitutes a sufficient explanation of what the AI system is doing? How should individuals be provided access to information such as the factors, the logic, and the techniques that produce an AI decision-making outcome?
- Fairness and Non-discrimination: e.g. What methods are available to detect and address potential unfairness in AI systems? How should AI developers mitigate bias in training data and in AI algorithms to avoid discriminatory impacts?
- Privacy: e.g. What types of privacy risks arise in the development and deployment of AI systems? How effective are data protection laws such as the General Data Protection Regulation (GDPR) at addressing those privacy risks?
The bridge will be a one-day meeting comprising invited talks from both AI and legal researchers; four breakout sessions to discuss the topics listed above; and a plenary panel where breakout session chairs will report a summary of their discussions for each topic, including but not limited to:
- What are open questions?
- What progress has been achieved in the last five years?
- What research agendas need to be pushed forward?
The meeting will end with a networking session.
- Bryan Choi (Ohio State University); email@example.com
- Lu Feng (University of Virginia); firstname.lastname@example.org
- Sarit Kraus (Bar Ilan University); email@example.com
- Christopher Yoo (University of Pennsylvania); firstname.lastname@example.org
Climate change has led to a rapid increase in the occurrence of extreme weather events globally, including floods, droughts, and heatwaves. Modeling climate is an essential endeavor to understand the near and long term impacts of climate change, as well as inform tools for mitigation and adaptation. However, first-principles models are limited in their ability to model complex dynamics that dictate weather and climate phenomena. With advancements in sensory technology and machine learning, AI can play a key role in advancing data-driven climate modeling with accurate and high-resolution forecasts of a variety of spatiotemporal phenomena related to weather and climate, such as agriculture yields, natural disasters, and socioeconomic welfare. In this bridge program, we aim to bring together students, researchers, and practitioners in both AI and climate science on topics related to interdisciplinary research and education in applying machine learning methods for climate science. Through an extensive program consisting of hands-on tutorials, paper presentations, and invited talks and panels with domain experts, we aim to familiarize the audience with the most pressing questions in climate science that can benefit from AI, the limitations of current data-driven models (e.g., incorporating physical constraints, quantifying uncertainty, distribution shifts), as well as educational resources and community initiatives for students, researchers, and practitioners in both disciplines.
The bridge will overview the role of AI in modeling various aspects of climate — atmospheric, oceanic, and land processes. For each of these topics, we will include discussions on the available data sources, the currently deployed methods, and (if any) AI approaches that have been explored in prior works. A key part of these discussions will be the limitations on both the data and modeling ends and potential solutions. Some of these questions include:
(a) What are the unique biases persisting existing datasets for climate science?
(b) How can we best incorporate physical domain knowledge in AI approaches?
(c) How can we model multimodal climate data spanning satellite imagery, simulation data, weather balloons, etc.?
(d) How robust are machine learning models for weather and climate across space and time?
(e) How can we quantify and calibrate uncertainties for forecasts with machine learning models?
(f) How can we scale AI infrastructure for training and inference to the petabyte-scale climate science datasets available today?
Our 1 day program covers a myriad of activities catering to students, researchers and practitioners interested in AI and climate science. Both the morning and afternoon talks will feature invited talks, contributed talks, and a panel discussion. Additionally, we will include a hands-on tutorial on machine learning for climate science and a poster session for all accepted papers.
The Bridge will be open to all participants in AAAI. Participants will be expected to have a basic background in machine learning. Knowledge of climate science is a plus, but not required.
We will be accepting electronic submissions between 4-8 pages. Submissions will be double blind on OpenReview. Further details and deadlines will be posted on the bridge website.
- Aditya Grover, UCLA, email@example.com
- Jayesh Gupta, Microsoft, firstname.lastname@example.org
- Ashish Kapoor, Microsoft, email@example.com
- Dev Niyogi, UT Austin, firstname.lastname@example.org
- Manmeet Singh, Indian Institute of Tropical Meteorology, email@example.com