Kenneth D. Forbus

AAMAS Conference 2025 Conference Paper

Reasoning and Planning with Dynamic Social Norms

• Taylor Olson
• Roberto Salas-Damian
• Kenneth D. Forbus

To safely interact with humans, AI systems must both have knowledge of our norms and consider norms in their planning processes. However, norm-guided planning has been less explored, only within communities of artificial agents and ignoring the dynamic nature of norms. This paper presents an approach to guiding planning with dynamically changing norms in a human-AI setting. This yields adaptive guard rails for the actions of AI systems.

IJCAI Conference 2024 Conference Paper

Normative Testimony and Belief Functions: A Formal Theory of Norm Learning

• Taylor Olson
• Kenneth D. Forbus

The ability to learn another’s moral beliefs is necessary for all social agents. It allows us to predict their behavior and is a prerequisite to correcting their beliefs if they are incorrect. To make AI systems more socially competent, a formal theory for learning internal normative beliefs is thus needed. However, to the best of our knowledge, a philosophically justified formal theory for this process does not yet exist. This paper begins the development of such a theory, focusing on learning from testimony. We make four main contributions. First, we provide a set of axioms that any such theory must satisfy. Second, we provide justification for belief functions, as opposed to traditional probability theory, for modeling norm learning. Third, we construct a novel learning function that satisfies these axioms. Fourth, we provide a complexity analysis of this formalism and proof that deontic rules are sound under its semantics. This paper thus serves as a theoretical contribution towards modeling learning norms from testimony, paving the road towards more social AI systems.

AAAI Conference 2023 Conference Paper

Mitigating Adversarial Norm Training with Moral Axioms

• Taylor Olson
• Kenneth D. Forbus

This paper addresses the issue of adversarial attacks on ethical AI systems. We investigate using moral axioms and rules of deontic logic in a norm learning framework to mitigate adversarial norm training. This model of moral intuition and construction provides AI systems with moral guard rails yet still allows for learning conventions. We evaluate our approach by drawing inspiration from a study commonly used in moral development research. This questionnaire aims to test an agent's ability to reason to moral conclusions despite opposed testimony. Our findings suggest that our model can still correctly evaluate moral situations and learn conventions in an adversarial training environment. We conclude that adding axiomatic moral prohibitions and deontic inference rules to a norm learning model makes it less vulnerable to adversarial attacks.

ICML Conference 2020 Conference Paper

Mapping natural-language problems to formal-language solutions using structured neural representations

• Kezhen Chen
• Qiuyuan Huang
• Hamid Palangi
• Paul Smolensky
• Kenneth D. Forbus
• Jianfeng Gao 0001

Generating formal-language programs represented by relational tuples, such as Lisp programs or mathematical operations, to solve problems stated in natural language is a challenging task because it requires explicitly capturing discrete symbolic structural information implicit in the input. However, most general neural sequence models do not explicitly capture such structural information, limiting their performance on these tasks. In this paper, we propose a new encoder-decoder model based on a structured neural representation, Tensor Product Representations (TPRs), for mapping Natural-language problems to Formal-language solutions, called TP-N2F. The encoder of TP-N2F employs TPR ‘binding’ to encode natural-language symbolic structure in vector space and the decoder uses TPR ‘unbinding’ to generate, in symbolic space, a sequential program represented by relational tuples, each consisting of a relation (or operation) and a number of arguments. TP-N2F considerably outperforms LSTM-based seq2seq models on two benchmarks and creates new state-of-the-art results. Ablation studies show that improvements can be attributed to the use of structured TPRs explicitly in both the encoder and decoder. Analysis of the learned structures shows how TPRs enhance the interpretability of TP-N2F.

AAAI Conference 2019 Conference Paper

Human-Like Sketch Object Recognition via Analogical Learning

• Kezhen Chen
• Irina Rabkina
• Matthew D. McLure
• Kenneth D. Forbus

Deep learning systems can perform well on some image recognition tasks. However, they have serious limitations, including requiring far more training data than humans do and being fooled by adversarial examples. By contrast, analogical learning over relational representations tends to be far more data-efficient, requiring only human-like amounts of training data. This paper introduces an approach that combines automatically constructed qualitative visual representations with analogical learning to tackle a hard computer vision problem, object recognition from sketches. Results from the MNIST dataset and a novel dataset, the Coloring Book Objects dataset, are provided. Comparison to existing approaches indicates that analogical generalization can be used to identify sketched objects from these datasets with several orders of magnitude fewer examples than deep learning systems require.

IS Journal 2017 Journal Article

Interactive Task Learning

• John E. Laird
• Kevin Gluck
• John Anderson
• Kenneth D. Forbus
• Odest Chadwicke Jenkins
• Christian Lebiere
• Dario Salvucci
• Matthias Scheutz

This article presents a new research area called interactive task learning (ITL), in which an agent actively tries to learn not just how to perform a task better but the actual definition of a task through natural interaction with a human instructor while attempting to perform the task. The authors provide an analysis of desiderata for ITL systems, a review of related work, and a discussion of possible application areas for ITL systems.

IJCAI Conference 2016 Conference Paper

Learning Paraphrase Identification with Structural Alignment

• Chen Liang
• Praveen Paritosh
• Vinodh Rajendran
• Kenneth D. Forbus

Semantic similarity of text plays an important role in many NLP tasks. It requires using both local information like lexical semantics and structural information like syntactic structures. Recent progress in word representation provides good resources for lexical semantics, and advances in natural language analysis tools make it possible to efficiently generate syntactic and semantic annotations. However, how to combine them to capture the semantics of text is still an open question. Here, we propose a new alignment-based approach to learn semantic similarity. It uses a hybrid representation, attributed relational graphs, to encode lexical, syntactic and semantic information. Alignment of two such graphs combines local and structural information to support similarity estimation. To improve alignment, we introduced structural constraints inspired by a cognitive theory of similarity and analogy. Usually only similarity labels are given in training data and the true alignments are unknown, so we address the learning problem using two approaches: alignment as feature extraction and alignment as latent variable. Our approach is evaluated on the paraphrase identification task and achieved results competitive with the state-of-the-art.

IJCAI Conference 2011 Conference Paper

Repairing Incorrect Knowledge with Model Formulation and Metareasoning

• Scott E. Friedman
• Kenneth D. Forbus

Learning concepts via instruction and expository texts is an important problem for modeling human learning and for making autonomous AI systems. This paper describes a computational model of the self-explanation effect, whereby conceptual knowledge is repaired by integrating and explaining new material. Our model represents conceptual knowledge with compositional model fragments, which are used to explain new material via model formulation. Preferences are computed over explanations and conceptual knowledge, along several dimensions. These preferences guide knowledge integration and question-answering. Our simulation learns about the human circulatory system, using facts from a circulatory system passage used in a previous cognitive psychology experiment. We analyze the simulation's performance, showing that individual differences in sequences of models learned by students can be explained by different parameter settings in our model.

IS Journal 2009 Journal Article

Companion Cognitive Systems: Design Goals and Lessons Learned So Far

• Kenneth D. Forbus
• Matthew Klenk
• Thomas Hinrichs

The companion cognitive architecture supports experiments in achieving human-level intelligence. This article describes seven key design goals of companions, relating them to properties of human reasoning and learning, and to engineering concerns raised by attempting to build large-scale cognitive systems. We summarize our experiences with companions in two domains: test taking and game playing.

IJCAI Conference 2007 Conference Paper

• Daniel T. Halstead
• Kenneth D. Forbus

A key issue in artificial intelligence lies in finding the amount of input detail needed to do successful learning. Too much detail causes overhead and makes learning prone to over-fitting. Too little detail and it may not be possible to learn anything at all. The issue is particularly relevant when the inputs are relational case descriptions, and a very expressive vocabulary may also lead to inconsistent representations. For example, in the Whodunit Problem, the task is to form hypotheses about the identity of the perpetrator of an event described using relational propositions. The training data consists of arbitrary relational descriptions of many other similar cases. In this paper, we examine the possibility of translating the case descriptions into an alternative vocabulary which has a reduced number of predicates and therefore produces more consistent case descriptions. We compare how the reduced vocabulary affects three different learning algorithms: exemplar-based analogy, prototype-based analogy, and association rule learning. We find that it has a positive effect on some algorithms and a negative effect on others, which gives us insight into all three algorithms and indicates when reduced vocabularies might be appropriate.

IJCAI Conference 2007 Conference Paper

• Thomas R. Hinrichs
• Kenneth D. Forbus

A key problem in playing strategy games is learning how to allocate resources effectively. This can be a difficult task for machine learning when the connections between actions and goal outputs are indirect and complex. We show how a combination of structural analogy, experimentation, and qualitative modeling can be used to improve performance in optimizing food production in a strategy game. Experimentation bootstraps a case library and drives variation, while analogical reasoning supports retrieval and transfer. A qualitative model serves as a partial domain theory to support adaptation and credit assignment. Together, these techniques can enable a system to learn the effects of its actions, the ranges of quantities, and to apply training in one city to other, structurally different cities. We describe experiments demonstrating this transfer of learning.

AAAI Conference 2007 Conference Paper

Integrating Natural Language, Knowledge Representation and Reasoning, and Analogical Processing to Learn by Reading

• Kenneth D. Forbus
• Lawrence Birnbaum
• Abhishek Sharma

Learning by reading requires integrating several strands of AI research. We describe a prototype system, Learning Reader, which combines natural language processing, a large-scale knowledge base, and analogical processing to learn by reading simplified language texts. We outline the architecture of Learning Reader and some of system-level results, then explain how these results arise from the components. Specifically, we describe the design, implementation, and performance characteristics of a natural language understanding model (DMAP) that is tightly coupled to a knowledge base three orders of magnitude larger than previous attempts. We show that knowing the kinds of questions being asked and what might be learned can help provide more relevant, efficient reasoning. Finally, we show that analogical processing provides a means of generating useful new questions and conjectures when the system ruminates off-line about what it has read.

AAAI Conference 2005 Conference Paper

Analogical Learning of Visual/Conceptual Relationships in Sketches

• Kenneth D. Forbus

This paper explores the use of analogy to learn about properties of sketches. Sketches often convey conceptual relationships between entities via the visual relationships between their depictions in the sketch. Understanding these conventions is an important part of adapting to a user. This paper describes how learning by accumulating examples can be used to make suggestions about such relationships in new sketches. We describe how sketches are being used in Companion Cognitive Systems to illustrate one context in which this problem arises. We describe how existing cognitive simulations of analogical matching and retrieval are used to generate suggestions for new sketches based on analogies with prior sketches. Two experiments provide evidence as to the accuracy and coverage of this technique.

AIJ Journal 1999 Journal Article

CyclePad: An articulate virtual laboratory for engineering thermodynamics

• Kenneth D. Forbus
• Peter B. Whalley
• John O. Everett
• Leo Ureel
• Mike Brokowski
• Julie Baher
• Sven E. Kuehne

One of the original motivations for research in qualitative physics was the development of intelligent tutoring systems and learning environments for physical domains and complex systems. This article demonstrates how a synergistic combination of qualitative reasoning and other AI techniques can be used to create an intelligent learning environment for students learning to analyze and design thermodynamic cycles. Pedagogically this problem is important because thermodynamic cycles express the key properties of systems which interconvert work and heat, such as power plants, propulsion systems, refrigerators, and heat pumps, and the study of thermodynamic cycles occupies a major portion of an engineering student's training in thermodynamics. This article describes CyclePad, a fully implemented articulate virtual laboratory that captures a substantial fraction of the knowledge in an introductory thermodynamics textbook and provides explanations of calculations and coaching support for students who are learning the principles of such cycles. CyclePad employs a distributed coaching model, where a combination of on-board facilities and a server-based coach accessed via email provide help for students, using a combination of teleological and case-based reasoning. CyclePad is a fielded system, in routine use in classrooms scattered all over the world. We analyze the combination of ideas that made CyclePad possible and comment on some lessons learned about the utility of various AI techniques based on our experience in fielding CyclePad.

AAAI Conference 1994 Conference Paper

Using Qualitative Physics to Build Articulate Software for Thermodynamics Education

• Kenneth D. Forbus

One of the original motivations for research in qualitative physics was the development of intelligent tutoring systems and learning environments for physical domains and complex systems. This paper demonstrates how a synergistic combination of qualitative physics and other AI techniques can be used to create an intelligent learning environment for students learning to analyze and design thermodynamic cycles. Pedagogically this problem is important because thermodynamic cycles express the key properties of systems which interconvert work and heat, such as power plants, propulsion systems, refrigerators, and heat pumps, and the study of thermodynamic cycles occupies a major portion of an engineering student' s training in thermodynamics. This paper describes CyclePad, a fully implemented learning environment which captures a substantial fraction of a thermodynamics textbook' s knowledge and is designed to scaffold students who are learning the principles of such cycles. We analyze the combination of ideas that made CyclePad possible, comment on some lessons learned about the utility of various techniques, and describe our plans for classroom experimentation.

AIJ Journal 1993 Journal Article

Qualitative process theory: twelve years after

• Kenneth D. Forbus

AAAI Conference 1992 Conference Paper

Self-Explanatory Simulations: Scaling Up to Large Models

• Kenneth D. Forbus

Qualitative reasoners have been hamstrung by the inability to analyze large models. This includes selfexplanatory simulators, which tightly integrate qualitative and numerical models to provide both precision and explanatory power. While they have important potential applications in training, instruction, and conceptual design, a critical step towards realizing this potential is the ability to build simulators for medium-sized systems (i. e. , on the order of ten to twenty independent parameters). This paper describes a new method for developing self-explanatory simulators which scales up. While our method involves qualitative analysis, it does not rely on envisioning or any other form of qualitative simulation. We describe the results of an implemented system which uses this method, and analyze its limitations and potential.

AIJ Journal 1991 Journal Article

Cmpositional modeling: finding the right model for the job

• Brian Falkenhainer
• Kenneth D. Forbus

To represent an engineer's knowledge will require domain theories that are orders of magnitude larger than today's theories, describe phenomena at several levels of granularity, and incorporate multiple perspectives. To build and use such theories effectively requires strategies for organizing domain models and techniques for determining which subset of knowledge to apply for a given task. This paper describes compositional modeling, a technique that addresses these issues. Compositional modeling uses explicit modeling assumptions to decompose domain knowledge into semi-independent model fragments, each describing various aspects of objects and physical processes. We describe an implemented algorithm for model composition. That is, given a general domain theory, a structural description of a specific system, and a query about the system's behavior, the algorithm composes a model which suffices to answer the query while minimizing extraneous detail. We illustrate the utility of compositional modeling by outlining the organization of a large-scale, multi-grain, multi-perspective model we have built for engineering thermodynamics, and showing how the model composition algorithm can be used to automatically select the appropriate knowledge to answer questions in a tutorial setting.

AIJ Journal 1991 Journal Article

Qualitative spatial reasoning: The CLOCK project

• Kenneth D. Forbus
• Paul Nielsen
• Boi Faltings

Spatial reasoning is ubiquitous in human problem solving. Significantly, many aspects of it appear to be qualitative. This paper describes a general framework for qualitative spatial reasoning and demonstrates how it can be used to understand complex mechanical systems, such as clocks. The framework is organized around three ideas. (1) We conjecture that no powerful, general-purpose, purely qualitative representation of spatial properties exists (the poverty conjecture). (2) We describe the MD/PV model of spatial reasoning, which overcomes this fundamental limitation by combining the power of diagrams with qualitative spatial representations. In particular, a metric diagram, which combines quantitative and symbolic information, is used as the foundation for constructing a place vocabulary, a symbolic representation of shape and space which supports qualitative spatial reasoning. (3) We claim that shape and connectivity are the central features of qualitative spatial representations for kinematics. We begin by exploring these ideas in detail, pointing out why simpler representations have not proven fruitful. We also describe how inferences can be organized using the MD/PV model. We demonstrate the utility of this model by describing clock, a program which reasons about complex two-dimensional mechanisms. clock starts with a CAD description of a mechanism's parts and constructs a qualitative simulation of how it can behave. clock successfully performed the first complete qualitative simulation of a mechanical clock from first principles, a milestone in qualitative physics. We also examine other work on qualitative spatial reasoning, and show how it fits into this framework. Finally, we discuss new research questions this framework raises.

AAAI Conference 1990 Conference Paper

Self-Explanatory Simulations: An Integration of Qualitative and Quantitative Knowledge

• Kenneth D. Forbus

A central goal of qualitative physics is to provide a framework for organizing and using quantitative knowledge. One important use of quantitative knowledge is numerical simulation. While current numerical simulators are powerful, they are often hard to construct, do not reveal the assumptions underlying their construction, and do not produce explanations of the behaviors they predict. This paper shows how to combine qualitative and quantitative models to produce a new class of self-explanatory simulations which combine the advantages of both kinds of reasoning. Self-explanat*ory simulations provide the accuracy of numerical models and the interpretive power of qualitative reasoning. We define what self-explanatory simulations are and show how to construct them automatically. We illustrate their power with some examples generated with an implemented system, SIHGEN. We analyze the limitations of our techniques, and discuss plans for future work.

IJCAI Conference 1989 Conference Paper

Introducing Actions into Qualitative Simulation

• Kenneth D. Forbus

AIJ Journal 1989 Journal Article

The structure-mapping engine: Algorithm and examples

• Brian Falkenhainer
• Kenneth D. Forbus
• Dedre Gentner

This paper describes the structure-mapping engine (SME), a program for studying analogical processing. SME has been built to explore Gentner's structure-mapping theory of analogy, and provides a “tool kit” for constructing matching algorithms consistent with this theory. Its flexibility enhances cognitive simulation studies by simplifying experimentation. Furthermore, SME is very efficient, making it a useful component in machine learning systems as well. We review the structure-mapping theory and describe the design of the engine. We analyze the complexity of the algorithm, and demonstrate that most of the steps are polynomial, typically bounded by O(N2). Next we demonstrate some examples of its operation taken from our cognitive simulation studies and work in machine learning. Finally, we compare SME to other analogy programs and discuss several areas for future work.

AAAI Conference 1988 Conference Paper

Focusing the ATMS

• Kenneth D. Forbus

IJCAI Conference 1987 Conference Paper

Qualitative Kinematics: A Framework

• Kenneth D. Forbus
• Paul Nielson
• Boi Faltings

Qualitative spatial reasoning has seen little progress This paper attempts to explain why We provide a framework for qualitative kinematics (QK), qualitative spatial reasoning about motion We propose that no general-purpose, purely qualitative kinematics exists. We propose instead the MD/PV model of spatial reasoning, which combines the power of diagrams with qualitative representations Next we propose connectivity as the organizing principle for kinematic state, and describe a set of basic inferences which every QK system must make. The framework's utility is illustrated by considering two programs, one finished and one in progress We end by discussing the research questions this framework raises.

IJCAI Conference 1987 Conference Paper

The Logic of Occurrence

• Kenneth D. Forbus

A general problem in qualitative physics is determin­ ing the consequences of assumptions about the be­ havior of a system. If the space of behaviors is repre­ sented by an envisionment, many such consequences can be represented by pruning states from the envi­ sionment. This paper provides a formal logic of occurrence which justifies the algorithms involved and provides a language for relating specific histories to envision ments The concepts and axioms are general enough to be applicable to any system of qualitative physics. We further propose the concept of transverse quantities as a general solution to qualitative versions of Zeno's paradox. The utility of these ideas is il­ lustrated by a rational reconstruction of the pruning algorithms used in FROB, a working AI program.

AAAI Conference 1986 Conference Paper

Interpreting Measurements of Physical Systems

• Kenneth D. Forbus

An unsolved problem in qualitative physics is generating a qualitative understanding of how a physical system is behaving from raw data, especially numerical data taken across time, to reveal changing internal state. Yet providing this ability to "read gauges" is a critical step towards building the next generation of intelligent computer-aided engineering systems and allowing robots to work in unconstrained envirionments. This paper presents a theory to solve this problem. Importantly, the theory is domain independent and will work with any system of qualitative physics. It requires only a qualitative description of the domain capable of supporting envisioning and domain-specific techniques for providing an initial qualitative description of numerical measurements. The theory has been fully implemented, and an extended example using Qualitative Process theory is presented.

AIJ Journal 1985 Journal Article

Structure and interpretation of computer programs

• Kenneth D. Forbus

AIJ Journal 1984 Journal Article

Qualitative process theory

• Kenneth D. Forbus

Objects move, collide, flow, bend, heat up, cool down, stretch, compress, and boil. These and other things that cause changes in objects over time are intuitively characterized as processes. To understand commonsense physical reasoning and make programs that interact with the physical world as well as people do we must understand qualitative reasoning about processes, when they will occur, their effects, and when they will stop. Qualitative process theory defines a simple notion of physical process that appears useful as a language in which to write dynamical theories. Reasoning about processes also motivates a new qualitative representation for quantity in terms of inequalities, called the quantity space. This paper describes the basic concepts of qualitative process theory, several different kinds of reasoning that can be performed with them, and discusses its implications for causal reasoning. Several extended examples illustrate the utility of the theory, including figuring out that a boiler can blow up, that an oscillator with friction will eventually stop, and how to say that you can pull with a string, but not push with it.

AAAI Conference 1982 Conference Paper

Modeling Motion with Qualitative Process Theory

• Kenneth D. Forbus

Representing motion is an important part of Naive Physics. Previous qualitative models of motion were centered around the idea of Qualitative States. This paper discusses an alternative representation in terms of Qualitative Process Theory. Viewing motion as a process has several advantages, notably the ability to make more detailed inferences about dynamics and the ability to combine process descriptions to model more complex systems. After examining the relationship between Qualitative State and QP theory representations of motion, the utility of the QP representations are illustrated by analyzing an oscillator.

AAAI Conference 1980 Conference Paper

Spatial and Qualitative Aspects of Reasoning about Motion

• Kenneth D. Forbus

Reasoning about motion is an important part of common sense knowledge. The spatial and qualitative aspects of reasoning about motion through fret space arc studied through the construction of a program to perform such reasoning. An analog gcomctry rcprcscntation serves as a diagram, and descriptions of both the actual motion of a ball and envisioning are used in answering simple questions.