Problems of autonomous agents following informal, open-textured rules

Quandt, Ryan; Licato, John

doi:10.1016/b978-0-12-820543-3.00011-0

Cited by 7 publications

(13 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In that example, the phrase 'impeding traffic' is opentextured (Hart 1961), meaning it refers to a category whose membership is "highly dependent on context and human intentions," where there is an "absence of precise conditions for membership" (Branting 2000). Although opentextured phrases introduce difficulties that are well-known (Sanders 1991;Bench-Capon 2012;Franklin 2012;Pereira et al 2017;Quandt and Licato 2019), their use is virtually unavoidable, even in domains as seemingly straightforward as traffic law (Prakken 2017). In practice, then, settling on an interpretation of open-textured phrases is done through argumentation.…”

Section: Interpretive Arguments Following the General Formulation Ofmentioning

confidence: 99%

“…Open-textured phrases are even more prevalent in ethical and legal domains. Whether they work fully autonomously or in human-machine teams, artificial agents given rules to follow (where those rules may range from international laws, to company ethical policies, to mission-specific orders) can benefit tremendously by understanding how to use interpretive reasoning to determine the applicability of opentextured phrases (Quandt and Licato 2019). For example, the ACM/IEE-CS Software Engineering Code of Ethics (Gotterbarn, Miller, and Rogerson 1997) states that software engineers should "[m]oderate the interests of the software engineer, the employer, the client and the users with the public good."…”

Section: Interpretive Arguments Following the General Formulation Ofmentioning

confidence: 99%

See 1 more Smart Citation

Scenarios and Recommendations for Ethical Interpretive AI

Licato,

Marji,

Abraham

2019

Preprint

Self Cite

View full text Add to dashboard Cite

Artificially intelligent systems, given a set of non-trivial ethical rules to follow, will inevitably be faced with scenarios which call into question the scope of those rules. In such cases, human reasoners typically will engage in interpretive reasoning, where interpretive arguments are used to support or attack claims that some rule should be understood a certain way. Artificially intelligent reasoners, however, currently lack the ability to carry out human-like interpretive reasoning, and we argue that bridging this gulf is of tremendous importance to human-centered AI. In order to better understand how future artificial reasoners capable of human-like interpretive reasoning must be developed, we have collected a dataset of ethical rules, scenarios designed to invoke interpretive reasoning, and interpretations of those scenarios. We perform a qualitative analysis of our dataset, and summarize our findings in the form of practical recommendations.

show abstract

Section: Interpretive Arguments Following the General Formulation Ofmentioning

confidence: 99%

Section: Interpretive Arguments Following the General Formulation Ofmentioning

confidence: 99%

Scenarios and Recommendations for Ethical Interpretive AI

Licato,

Marji,

Abraham

2019

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Lawmakers often intentionally use vague language, in part because it allows for the delegation of discretion to boots-on-the-ground agents [2,62]. 1 Insofar as they provide the means for flexibility in interpretation, open-textured terms are an unavoidable and necessary feature of legal, ethical, and policy regulations [19,27,35,36,54,56]. As such, autonomous reasoning software systems at all levels must learn how to work with them.…”

Section: Introduction: Why Addressing Open-texturedness Mattersmentioning

confidence: 99%

“…In short: rules that are too formal and specific tend to be difficult to understand / communicate, and tend to have such a narrow domain of applicability as to make them of little use in complex domains. It is likely impossible that rules in any representational system, when those rules must govern behavior in complex real-world domains, and the rules must have a reasonable degree of human-understandability, can completely avoid open-textured concepts (and complete formal rigidity in such rules may not be preferable anyway [1,22,35,55]). Consider the language used in rules in ethical and legal domains.…”

Section: Introduction: Why Addressing Open-texturedness Mattersmentioning

confidence: 99%

“…Consider the language used in rules in ethical and legal domains. Whether they work fully autonomously or in human-machine teams, artificial agents given rules to follow (for rules ranging from international laws, to company ethical policies, to mission-specific orders) can benefit tremendously by understanding how to use interpretive reasoning to determine the applicability of open-textured phrases [55]. For example, the ACM/IEE-CS Software Engineering Code of Ethics [20] states that software engineers should "[m]oderate the interests of the software engineer, the employer, the client and the users with the public good."…”

Section: Introduction: Why Addressing Open-texturedness Mattersmentioning

confidence: 99%

See 1 more Smart Citation

How Should AI Interpret Rules? A Defense of Minimally Defeasible Interpretive Argumentation

Licato¹

2021

Preprint

Self Cite

View full text Add to dashboard Cite

Can artificially intelligent systems follow rules? The answer might seem an obvious 'yes', in the sense that all (current) AI strictly acts in accordance with programming code constructed from highly formalized and well-defined rulesets. But here I refer to the kinds of rules expressed in human language that are the basis of laws, regulations, codes of conduct, ethical guidelines, and so on. The ability to follow such rules, and to reason about them, is not nearly as clear-cut as it seems on first analysis. Real-world rules are unavoidably rife with open-textured terms, which imbue rules with a possibly infinite set of possible interpretations. Narrowing down this set requires a complex reasoning process that is not yet within the scope of contemporary AI.This poses a serious problem for autonomous AI: If one cannot reason about open-textured terms, then one cannot reason about (or in accordance with) real-world rules. And if one cannot reason about real-world rules, then one cannot: follow human laws, comply with regulations, act in accordance with written agreements, or even obey mission-specific commands that are anything more than trivial. But before tackling these problems, we must first answer a more fundamental question: Given an open-textured rule, what is its correct interpretation? Or more precisely: How should our artificially intelligent systems determine which interpretation to consider correct? In this essay, I defend the following answer: Rule-following AI should act in accordance with the interpretation best supported by minimally defeasible interpretive arguments (MDIA). In Section 2, I will introduce and defend the need for AI which can produce interpretations of rules that are supported by interpretive argumentation, and advocate for AI that is argument-justified rather than merely explainable. But in order to recognize interpretive argument quality, we need a way to estimate argument strength. Section 3 argues that the minimal defeasibility standard is our best hope. Finally, I address anticipated counterarguments and possible points of confusion in Section 4.

show abstract