Causation in AI and Law

Reasoning about causation in fact is an essential element of attributing legal responsibility. Therefore, the automation of the attribution of legal responsibility requires a modelling effort aimed at the following: a thorough understanding of the relation between the legal concepts of responsibility and of causation in fact; a thorough understanding of the relation between causation in fact and the common sense concept of causation; and, finally, the specification of an ontology of the concepts that are minimally required for (automatic) common sense reasoning about causation. This article offers a worked-out example of the indicated analysis. Such example consists of: a definition of the legal concept of responsibility (in terms of liability and accountability); a definition of the legal concept of causation in fact (in terms of the initiation of physical processes by an agent and of the provision of reasons and/or opportunities to other agents); CausatiOnt, an AI-like ontology of the common sense (causal) concepts that are minimally needed for reasoning about the legal concept of causation in fact (in particular, the concepts of category, dimension, object, agent, process, event and act).     

Hitchcock’s (2001) treatment of singular and general causation

Hitchcock (2001a) argues that the distinction between singular and general causation conflates the two distinctions ‘actual causation vs. causal tendencies’ and ‘wide vs. narrow causation’. Based on a recent regularity account of causation I will show that Hitchcock’s introduction of the two distinctions is an unnecessary multiplication of causal concepts.

The Overdetermination Argument Revisited

In this paper I discuss a famous argument for physicalism – which some authors indeed regard as the only argument for it – the overdetermination argument. In fact it is an argument that does not establish that all the entities in the world are physical, but that all those events that enter into causal transactions with the physical world are physical. As mental events seem to cause changes in the physical world, the mind is one of those things that fall within the scope of the argument. Here I analyze one response to the overdetermination argument that has acquired some popularity lately, and which consists in saying that what mental events cause are not physical effects. I try to show that recent attempts to develop this response are not successful, but that there may be a coherent way of doing so. I also try to show that there seems to be a philosophical niche in which this way might fit.     

Qualitative and quantitative conditions for the transitivity of perceived causation:

If A caused B and B caused C, did A cause C? Although laypersons commonly perceive causality as being transitive, some philosophers have questioned this assumption, and models of causality in artificial intelligence are often agnostic with respect to transitivity. We consider two formal models of causation that differ in the way they represent uncertainty. The quantitative model uses a crude probabilistic definition, arguably the common core of more sophisticated quantitative definitions; the qualitative model uses a definition based on nonmonotonic consequence relations. Different sufficient conditions for the transitivity of causation are laid bare by the two models: The Markov condition on events for the quantitative model, and a so-called saliency condition (A is perceived as a typical cause of B) for the qualitative model. We explore the formal and empirical relations between these sufficient conditions, and between the underlying definitions of perceived causation. These connections shed light on the range of applicability of each model, contrasting commonsense causal reasoning (supposedly qualitative) and scientific causation (more naturally quantitative). These speculations are supported by a series of three behavioral experiments.     

Reading direction and signaling in a simple computer simulation

The experiment investigated whether layout of cause and effect affects learning for causal connections in a simple computer simulation. Students (N = 113) used an introductory text and a simulation to learn central concepts about neural networks and then took a retention and transfer test. Each learner was randomly assigned to one cell of a 2 (compliance with reading direction or not) × 2 (with or without signaling) between subjects factorial design. Students who obtained the causal connection in reading direction (the cause is positioned on the left side, the effect is placed on the right side) performed better on transfer than did students, for whom the reading direction was reversed (from right to left). Furthermore, signals that indicated the layout of the causal connection fostered transfer performance and reduced time spent with the simulation. These results are consistent with the signaling principle and the cognitive theory of multimedia learning.     

Learning Causes: Psychological Explanations of Causal Explanation1

I argue that psychologists interested in human causal judgment should understand and adopt a representation of causal mechanisms by directed graphs that encode conditional independence (screening off) relations. I illustrate the benefits of that representation, now widely used in computer science and increasingly in statistics, by (i) showing that a dispute in psychology between mechanist and associationist psychological theories of causation rests on a false and confused dichotomy; (ii) showing that a recent, much-cited experiment, purporting to show that human subjects, incorrectly let large causes overshadow small causes, misrepresents the most likely, and warranted, causal explanation available to the subjects, in the light of which their responses were normative; (iii) showing how a recent psychological theory (due to P. Cheng) of human judgment of causal power can be considerably generalized: and (iv) suggesting a range of possible experiments comparing human and computer abilities to extract causal information from associations.     

Correlation between Different Hand Force Assessment Methods from an Epidemiological Study

This article presents the outcome of correlation analyses of data results obtained from using different methods for objectively and subjectively assessing hand force from a prospective study of 450 blue‐collar workers from several companies and industries, followed for up to 3 years. The study collected detailed ergonomic exposure data at baseline and upper extremity health outcome data at baseline and every 4 months during the study. Ultimately, the study was intended to evaluate dose‐response relationships of specific upper extremity disorders with detailed physical and psychosocial exposure data at the workplace while controlling for important individual factors. This article presents the methods used to collect data, as well as the hand force results of the epidemiological study in aggregate correlated form, as a means of exploring the degree of independence between the variables considered. These insights are useful in identifying musculoskeletal disorder (MSD) causation and predicting MSD risk based on work exposures. An enhanced understanding of the independence of MSD causal factors is instrumental in establishing more accurate multivariable models of MSD causation that will play an important role in extrapolating from the understanding of mechanisms of causation to establishing effective recommendations and programs to prevent the occurrence of MSDs. © 2011 Wiley Periodicals, Inc.     

Detection of Unfaithfulness and Robust Causal Inference

Much of the recent work on the epistemology of causation has centered on two assumptions, known as the Causal Markov Condition and the Causal Faithfulness Condition. Philosophical discussions of the latter condition have exhibited situations in which it is likely to fail. This paper studies the Causal Faithfulness Condition as a conjunction of weaker conditions. We show that some of the weaker conjuncts can be empirically tested, and hence do not have to be assumed a priori. Our results lead to two methodologically significant observations: (1) some common types of counterexamples to the Faithfulness condition constitute objections only to the empirically testable part of the condition; and (2) some common defenses of the Faithfulness condition do not provide justification or evidence for the testable parts of the condition. It is thus worthwhile to study the possibility of reliable causal inference under weaker Faithfulness conditions. As it turns out, the modification needed to make standard procedures work under a weaker version of the Faithfulness condition also has the practical effect of making them more robust when the standard Faithfulness condition actually holds. This, we argue, is related to the possibility of controlling error probabilities with finite sample size (“uniform consistency”) in causal inference.

Causality, conditional independence, and graphical separation in settable systems

We study the connections between causal relations and conditional independence within the settable systems extension of the Pearl causal model (PCM). Our analysis clearly distinguishes between causal notions and probabilistic notions, and it does not formally rely on graphical representations. As a foundation, we provide definitions in terms of suitable functional dependence for direct causality and for indirect and total causality via and exclusive of a set of variables. Based on these foundations, we provide causal and stochastic conditions formally characterizing conditional dependence among random vectors of interest in structural systems by stating and proving the conditional Reichenbach principle of common cause, obtaining the classical Reichenbach principle as a corollary. We apply the conditional Reichenbach principle to show that the useful tools of d-separation and D-separation can be employed to establish conditional independence within suitably restricted settable systems analogous to Markovian PCMs.     

A contrarian view of postmodern society and information technologies

In this short paper—little more than a note, even a short “contrarian” sermon for this anniversary volume—what I do is argue that even the allegedly most “revolutionary” inventions of our computer-driven age are not revolutionary in the sense that their impacts are “driving” society. Some of them are genuinely revolutionary, I admit, but in the reverse direction. The inventions don’t “impact societies”; rather, particular communities within society use the technical languages that are at their core, invent them, embed them in machines, and so on. It is not inventions but particular groups within modern—and so-called postmodern—societies that have invented and use technical languages which are embedded in gadgets that are said to “drive” modern or postmodern societies. And they do so only in one sense: they were invented and are used by various communities in our kinds of societies for a variety of ends. And if this is so, and if we feel those ends are undemocratic or positively anti-democratic, I conclude that we should resist them any way we can, even politically.     

基于模糊理论的煤矿供电系统接地故障选线方法

针对单一接地故障选线方法经常出现选择性失效的问题,提出了一种将零序全电流功率方向原理、小波变换分析原理、灰色关联分析原理相结合的接地故障选线方法。该方法利用零序电压多重分形谱确定方法权重,使用模糊推理理论进行多判据信息融合。Matlab仿真实验结果表明,通过定位方法确定权重的复合判据适用于各种情况下的单相接地故障选线,不受系统中性点接地方式、运行方式、出线路数及接地位置的影响,实现了高准确率的单相接地故障定位。     

The Causal and Explanatory Role of Information Stored in Connectionist Networks

In this paper I defend the propriety of explaining the behavior of distributed connectionist networks by appeal to selected data stored therein. In particular, I argue that if there is a problem with such explanations, it is a consequence of the fact that information storage in networks is superpositional, and not because it is distributed. I then develop a ``proto-account' of causation for networks, based on an account of Andy Clark's, that shows even superpositionality does not undermine information-based explanation. Finally, I argue that the resulting explanations are genuinely informative and not vacuous.     

Is There Life After the Death of the Computational Theory of Mind?

In this paper, I explore the implications of Fodor’s attacks on the Computational Theory of Mind (CTM), which get their most recent airing in The Mind Doesn’t Work That Way. I argue that if Fodor is right that the CTM founders on the global nature of abductive inference, then several of the philosophical views about the mind that he has championed over the years founder as well. I focus on Fodor’s accounts of mental causation, psychological explanation, and intentionality.     

Design representations of moving bodies for interactive,motion-sensing spaces

This paper describes the development and use of a set of design representations of moving bodies in the design of Bystander, a multi-user, interactive, immersive artwork built on video-based, motion-sensing technology. We extended the traditional user-centred design tools of personas and scenarios to explicitly address human movement characteristics embedded in social interaction. A set of corresponding movement schemas in Labanotation was constructed to visually represent the spatial and social interaction of multiple users over time. Together these three design representations of moving bodies were used to enable the design team to work with the aspects of human movement relevant to Bystander and to ensure that the system could respond in a coherent and robust manner to the shifting configurations of visitors in the space. They also supported two experiential methods of design reflection-in-action—enactment and immersion—that were vital for grounding designers’ understandings of the specific interactive nature of the work in their own sensing, feeling and moving bodies.     

A Misdirected Principle with a Catch: Explicability for AI

There is widespread agreement that there should be a principle requiring that artificial intelligence (AI) be ‘explicable’. Microsoft, Google, the World Economic Forum, the draft AI ethics guidelines for the EU commission, etc. all include a principle for AI that falls under the umbrella of ‘explicability’. Roughly, the principle states that “for AI to promote and not constrain human autonomy, our ‘decision about who should decide’ must be informed by knowledge of how AI would act instead of us” (Floridi et al. in Minds Mach 28(4):689–707, 2018). There is a strong intuition that if an algorithm decides, for example, whether to give someone a loan, then that algorithm should be explicable. I argue here, however, that such a principle is misdirected. The property of requiring explicability should attach to a particular action or decision rather than the entity making that decision. It is the context and the potential harm resulting from decisions that drive the moral need for explicability—not the process by which decisions are reached. Related to this is the fact that AI is used for many low-risk purposes for which it would be unnecessary to require that it be explicable. A principle requiring explicability would prevent us from reaping the benefits of AI used in these situations. Finally, the explanations given by explicable AI are only fruitful if we already know which considerations are acceptable for the decision at hand. If we already have these considerations, then there is no need to use contemporary AI algorithms because standard automation would be available. In other words, a principle of explicability for AI makes the use of AI redundant.

     

Adaptive persuasive messaging to increase service retention: using persuasion profiles to increase the effectiveness of email reminders

In this article, we describe the usage of persuasion profiles in a large scale, N = 1,129, field trial. Persuasive technologies—technologies intentionally designed to influence user behavior—are emergent and becoming more and more individualized and ubiquitous. Individual differences in people’s responses to often used persuasion principles—different psychological means by which to influence users—motivate personalization. We describe how, through identification, representation, and measurement, persuasive technologies can personalize their persuasive attempts. Next, we show that dynamically adapting a persuasive technology to the responses of its users increases the effectiveness of the system. Ubiquitous computing systems are, because of their ability to unobtrusively measure user behavior, very well suited for these types of applications.     

Boosting scene understanding by hierarchical pachinko allocation

Scene understanding is a popular research direction. In this area, many attempts focus on the problem of naming objects in the complex natural scene, and visual semantic integration model (VSIM) is the representative. This model consists of two parts: semantic level and visual level. In the first level, it uses a four-level pachinko allocation model (PAM) to capture the semantics behind images. However, this four-level PAM is inflexible and lacks of considerations of common subtopics that represent the background semantics. To address these problems, we use hierarchical PAM (hPAM) to replace PAM. Since hPAM is flexible, we investigate two variations of hPAM to boost VSIM in this paper. We derive the Gibbs sampler to learn the proposed models. Empirical results validate that our works can obtain better performance than the state-of-the-art algorithms.     

Geometric mouldability analysis by geometric reasoning and fuzzy decision making

This paper presents a methodology for mouldability analysis by finding the optimal cavity design scheme (CDS) based on manufacturing and cost considerations using part geometry, where a CDS refers to a combination of the parting direction, parting line (PL), and undercut features (UF). The methodology takes advantage of geometric reasoning and fuzzy evaluation, and consists of two main stages: (1) generating all possible design alternatives, and (2) choosing the best alternative. In the first stage, after recognizing the potential UF from the given part, a spherical arrangement is constructed by partitioning the unit direction sphere using outward normals of the part's surfaces with the property that each cell in this arrangement has a unique combination of PL and UF set. Thus all design alternatives can be identified in O(ml²) time by visiting the cells in a certain order and updating the PL and UF set incrementally, where m and l are the number of the part's convex and overall surfaces, respectively. In the second stage, the fuzzy multiple attribute decision-making model is employed to choose the optimal scheme from the set of design alternatives with respect to a set of criteria related to the number and volume of undercuts, flatness of the PL, draw depth, and draft angle. This model allows designers to describe their preferences on different criteria in imprecise linguistic statements. Finally, the case studies show that the proposed methodology is very effective in finding the optimal CDS for the molded part and the final results conform to human designers' experience.