Practical Bandits: An Industry Perspective

Bram van den Akker (Booking), Olivier Jeunen (ShareChat), Ying Li (Netflix), Ben London (Amazon), Zahra Nazari (Spotify) and Devesh Parekh (Netflix)

The bandit paradigm provides a unified modeling framework for problems that require decision-making under uncertainty. Because many business metrics can be viewed as rewards (a.k.a. utilities) that result from actions, bandit algorithms have seen a large and growing interest from industrial applications, such as search, recommendation and advertising. Indeed, with the bandit lens comes the promise of direct optimization for the metrics we care about. Nevertheless, the road to successfully applying bandits in production is not an easy one. Even when the action space and rewards are well-defined, practitioners still need to make decisions regarding multi-arm or contextual approaches, on- or off-policy setups, delayed or immediate feedback, myopic or long-term optimisation, etc. To make matters worse, industrial platforms typically give rise to large action spaces in which existing approaches tend to break down. The research literature on these topics is broad and vast, but this can overwhelm practitioners, whose primary aim is to solve practical problems, and therefore need to decide on a specific instantiation or approach for each project. This tutorial will take a step towards filling that gap between the theory and practice of bandits. Our goal is to present a unified overview of the field and its existing terminology, concepts and algorithms—with a focus on problems relevant to industry. We hope our industrial perspective will help future practitioners who wish to leverage the bandit paradigm for their application.

Some Useful Things to Know When Combining IR and NLP: the Easy, the Hard and the Ugly

Omar Alonso (Amazon) and Kenneth Church (Northeastern University)

Deep nets such as GPT are at the core of the current advances in many systems and applications. Things are moving fast; techniques become obsolete quickly (within weeks). How can we take advantage of new discoveries and incorporate them into our existing work? Are new developments radical improvements, or incremental repetitions of established concepts, or combinations of both? In this tutorial, we aim to bring interested researchers and practitioners up to speed on the recent and ongoing techniques around ML and Deep learning in the context of IR and NLP. Additionally, our goal is to clarify terminology, emphasize fundamentals, and outline problems and new research opportunities.

Bridging Text Data and Graph Data: Towards Semantics and Structure-aware Knowledge Discovery

Bowen Jin, Yu Zhang, Sha Li and Jiawei Han (UIUC)

Graphs and texts play crucial roles in data mining, each possessing unique characteristics that often require distinct modeling methods. Technologies for mining graph data and text data are usually designed separately. Nevertheless, frequently, data contains a blend of both modalities, with their information frequently complementing each other. For instance, in e-commerce data, the product-user graph and product descriptions provide distinct insights into product features. Similarly, in scientific literature, the citation graph, author information, and the content of papers collectively contribute to modeling the impact of a paper.
In this tutorial, our emphasis will be on exploring the latest advancements in graph mining techniques that leverage the capabilities of Pre-trained Language Models (PLMs), as well as the enhancement of text mining methods through the incorporation of graph structure information. We will present an organized picture of how graphs and texts can mutually benefit each other and lead to deeper knowledge discovery, with the following outline: (1) an introduction to how graph and text are intertwined in real-life data and how graph neural networks and pre-trained language models are designed to capture signal from graph and text modalities, (2) graph construction from text: construct sentence-level graphs, event graphs, reasoning, knowledge graphs from text, (3) network mining with language models: language model-based methods for representation learning on graph and language model pretraining on graphs, (4) text mining with structure information: text classification, literature retrieval, and question answering with graph structure as auxiliary information, (5) towards an integrated semantics and structure mining paradigm.

Unbiased Learning to Rank: On Recent Advances and Practical Applications

Shashank Gupta (UvA), Philipp Hager (UvA), Jin Huang (UvA), Ali Vardasbi (UvA) and Harrie Oosterhuis (Radboud University)

Since its inception, the field of unbiased learning to rank (ULTR) has remained very active and has seen several impactful advancements in recent years. This tutorial provides both an introduction to the core concepts of the field and an overview of recent advancements in its foundations, along with several applications of its methods. The tutorial is divided into four parts: Firstly, we give an overview of the different forms of bias that can be addressed with ULTR methods. Secondly, we present a comprehensive discussion of the latest estimation techniques in the ULTR field.Thirdly, we survey published results of ULTR in real-world applications. Fourthly, we discuss the connection between ULTR and fairness in ranking. We end by briefly reflecting on the future of ULTR research and its applications. This tutorial is intended to benefit both researchers and industry practitioners interested in developing new ULTR solutions or utilizing them in real-world applications.

Strategic ML: How to Learn With Data That ‘Behaves’

Nir Rosenfeld (Technion)

The success of machine learning across a wide array of tasks and applications has made it appealing to use it also in the social domain. Indeed, learned models now form the backbone of recommendation systems, social media platforms, online markets, and e-commerce services, where they are routinely used to inform decisions by, for, and about their human users. But humans are not your conventional input–they have goals, beliefs, and aspirations, and take action to promote their own interests. Given that standard learning methods are not designed to handle inputs that `behave’, a natural question is: how should we design learning systems when we know they will be deployed and used in social settings? This tutorial introduces strategic machine learning, a new and emerging subfield of machine learning that aims to develop a disciplined framework for learning under strategic user behavior. The working hypothesis of strategic ML is simple: users want things, and act to achieve them. Surprisingly, this basic truism is difficult to address within the conventional learning framework. The key challenge is that how users behave often depends on the learned decision rule itself; thus, strategic learning seeks to devise methods which are able to anticipate and accommodate such responsive behavior. Towards this, strategic ML offers a formalism for reasoning about strategic responses, for designing appropriate learning objectives, and for developing practical tools for learning in strategic environments. The tutorial will survey recent and ongoing work in this new domain, present key theoretical and empirical results, provide practical tools, and discuss open questions and landmark challenges.

Responsible AI

Ricardo Baeza Yates (Northeastern University)

In the first part of this tutorial we define responsible AI and we discuss the problems embedded in terms like ethical or trustworthy AI. In the second part, to set the stage, we cover irresponsible AI: discrimination (e.g., the impact of human biases); pseudo-science (e.g., biometric based behavioral predictions); human limitations (e.g., human incompetence, cognitive biases); technical limitations (data as a proxy of reality, wrong evaluation); social impact (e.g., unfair digital markets or mental health and disinformation issues created by large language models); environmental impact (e.g., indiscriminate use of computing resources). These examples do have a personal bias but set the context for the third part where we cover the current challenges: ethical principles, governance and regulation. We finish by discussing our responsible AI initiatives, many recommendations, and some philosophical issues.

Trustworthy LLMs

Sanmi Koyejo (Stanford University) and Bo Li (U Chicago)

Large Language models are among the most exciting technologies developed in the last few years. While the model’s capabilities continue to improve, researchers, practitioners, and the general public are increasingly aware of some of its shortcomings. What will it take to build trustworthy large language models? This tutorial will present a range of recent findings, discussions, questions, and partial answers in the space of trustworthiness in large language models. While this tutorial will not attempt a comprehensive overview of this rich area, we aim to provide the participants with some tools and insights and to understand both the conceptual foundations of trustworthiness and a broad range of ongoing research efforts. We will tackle some of the hard questions that you may have about trustworthy large language models and hopefully address some misconceptions that have become pervasive.