Handling Reward Misspecification in the Presence of Expectation Mismatch

Detecting and handling misspecified objectives, such as reward functions, has been widely recognized as one of the central challenges within the domain of Artificial Intelligence (AI) safety research. However, even with the recognition of the importance of this problem, we are unaware of any works that attempt to provide a clear definition for what constitutes (a) misspecified objectives and (b) successfully resolving such misspecifications. In this work, we use the theory of mind, i.e., the human user's beliefs about the AI agent, as a basis to develop a formal explanatory framework called Expectation Alignment (EAL) to understand the objective misspecification and its causes. Our \EAL\ framework not only acts as an explanatory framework for existing works but also provides us with concrete insights into the limitations of existing methods to handle reward misspecification and novel solution strategies. We use these insights to propose a new interactive algorithm that uses the specified reward to infer potential user expectations about the system behavior. We show how one can efficiently implement this algorithm by mapping the inference problem into linear programs. We evaluate our method on a set of standard Markov Decision Process (MDP) benchmarks.

Towards More Likely Models for AI Planning

This is the first work to look at the application of large language models (LLMs) for the purpose of model space edits in automated planning tasks. To set the stage for this sangam, we explore two different flavors of model space problems that have been studied in the AI planning literature and explore the effect of an LLM on those tasks. We empirically demonstrate how the performance of an LLM contrasts with combinatorial search (CS) - an approach that has been traditionally used to solve model space tasks in planning, both with the LLM in the role of a standalone model space reasoner as well as in the role of a statistical signal in concert with the CS approach as part of a two-stage process. Our experiments show promising results suggesting further forays of LLMs into the exciting world of model space reasoning for planning tasks in the future.

TOBY: A Tool for Exploring Data in Academic Survey Papers

This paper describes TOBY, a visualization tool that helps a user explore the contents of an academic survey paper. The visualization consists of four components: a hierarchical view of taxonomic data in the survey, a document similarity view in the space of taxonomic classes, a network view of citations, and a new paper recommendation tool. In this paper, we will discuss these features in the context of three separate deployments of the tool.

On the Planning Abilities of Large Language Models -- A Critical Investigation

Intrigued by the claims of emergent reasoning capabilities in LLMs trained on general web corpora, in this paper, we set out to investigate their planning capabilities. We aim to evaluate (1) the effectiveness of LLMs in generating plans autonomously in commonsense planning tasks and (2) the potential of LLMs as a source of heuristic guidance for other agents (AI planners) in their planning tasks. We conduct a systematic study by generating a suite of instances on domains similar to the ones employed in the International Planning Competition and evaluate LLMs in two distinct modes: autonomous and heuristic. Our findings reveal that LLMs' ability to generate executable plans autonomously is rather limited, with the best model (GPT-4) having an average success rate of ~12% across the domains. However, the results in the heuristic mode show more promise. In the heuristic mode, we demonstrate that LLM-generated plans can improve the search process for underlying sound planners and additionally show that external verifiers can help provide feedback on the generated plans and back-prompt the LLM for better plan generation.

Leveraging Pre-trained Large Language Models to Construct and Utilize World Models for Model-based Task Planning

There is a growing interest in applying pre-trained large language models (LLMs) to planning problems. However, methods that use LLMs directly as planners are currently impractical due to several factors, including limited correctness of plans, strong reliance on feedback from interactions with simulators or even the actual environment, and the inefficiency in utilizing human feedback. In this work, we introduce a novel alternative paradigm that constructs an explicit world (domain) model in planning domain definition language (PDDL) and then uses it to plan with sound domain-independent planners. To address the fact that LLMs may not generate a fully functional PDDL model initially, we employ LLMs as an interface between PDDL and sources of corrective feedback, such as PDDL validators and humans. For users who lack a background in PDDL, we show that LLMs can translate PDDL into natural language and effectively encode corrective feedback back to the underlying domain model. Our framework not only enjoys the correctness guarantee offered by the external planners but also reduces human involvement by allowing users to correct domain models at the beginning, rather than inspecting and correcting (through interactive prompting) every generated plan as in previous work. On two IPC domains and a Household domain that is more complicated than commonly used benchmarks such as ALFWorld, we demonstrate that GPT-4 can be leveraged to produce high-quality PDDL models for over 40 actions, and the corrected PDDL models are then used to successfully solve 48 challenging planning tasks. Resources including the source code will be released at: https://guansuns.github.io/pages/llm-dm.

Planning for Attacker Entrapment in Adversarial Settings

In this paper, we propose a planning framework to generate a defense strategy against an attacker who is working in an environment where a defender can operate without the attacker's knowledge. The objective of the defender is to covertly guide the attacker to a trap state from which the attacker cannot achieve their goal. Further, the defender is constrained to achieve its goal within K number of steps, where K is calculated as a pessimistic lower bound within which the attacker is unlikely to suspect a threat in the environment. Such a defense strategy is highly useful in real world systems like honeypots or honeynets, where an unsuspecting attacker interacts with a simulated production system while assuming it is the actual production system. Typically, the interaction between an attacker and a defender is captured using game theoretic frameworks. Our problem formulation allows us to capture it as a much simpler infinite horizon discounted MDP, in which the optimal policy for the MDP gives the defender's strategy against the actions of the attacker. Through empirical evaluation, we show the merits of our problem formulation.

On the Planning Abilities of Large Language Models (A Critical Investigation with a Proposed Benchmark)

Intrigued by the claims of emergent reasoning capabilities in LLMs trained on general web corpora, in this paper, we set out to investigate their planning capabilities. We aim to evaluate (1) how good LLMs are by themselves in generating and validating simple plans in commonsense planning tasks (of the type that humans are generally quite good at) and (2) how good LLMs are in being a source of heuristic guidance for other agents--either AI planners or human planners--in their planning tasks. To investigate these questions in a systematic rather than anecdotal manner, we start by developing a benchmark suite based on the kinds of domains employed in the International Planning Competition. On this benchmark, we evaluate LLMs in three modes: autonomous, heuristic and human-in-the-loop. Our results show that LLM's ability to autonomously generate executable plans is quite meager, averaging only about 3% success rate. The heuristic and human-in-the-loop modes show slightly more promise. In addition to these results, we also make our benchmark and evaluation tools available to support investigations by research community.

Goal Alignment: A Human-Aware Account of Value Alignment Problem

Value alignment problems arise in scenarios where the specified objectives of an AI agent don't match the true underlying objective of its users. The problem has been widely argued to be one of the central safety problems in AI. Unfortunately, most existing works in value alignment tend to focus on issues that are primarily related to the fact that reward functions are an unintuitive mechanism to specify objectives. However, the complexity of the objective specification mechanism is just one of many reasons why the user may have misspecified their objective. A foundational cause for misalignment that is being overlooked by these works is the inherent asymmetry in human expectations about the agent's behavior and the behavior generated by the agent for the specified objective. To address this lacuna, we propose a novel formulation for the value alignment problem, named goal alignment that focuses on a few central challenges related to value alignment. In doing so, we bridge the currently disparate research areas of value alignment and human-aware planning. Additionally, we propose a first-of-its-kind interactive algorithm that is capable of using information generated under incorrect beliefs about the agent, to determine the true underlying goal of the user.

A Mental Model Based Theory of Trust

Handling trust is one of the core requirements for facilitating effective interaction between the human and the AI agent. Thus, any decision-making framework designed to work with humans must possess the ability to estimate and leverage human trust. In this paper, we propose a mental model based theory of trust that not only can be used to infer trust, thus providing an alternative to psychological or behavioral trust inference methods, but also can be used as a foundation for any trust-aware decision-making frameworks. First, we introduce what trust means according to our theory and then use the theory to define trust evolution, human reliance and decision making, and a formalization of the appropriate level of trust in the agent. Using human subject studies, we compare our theory against one of the most common trust scales (Muir scale) to evaluate 1) whether the observations from the human studies match our proposed theory and 2) what aspects of trust are more aligned with our proposed theory.