Why should moral philosophers, moral psychologists, and machine ethicists care about computational complexity? Debates on whether artificial intelligence (AI) can or should be used to solve problems in ethical domains have mainly been driven by what AI can or cannot do in terms of human capacities. In this paper, we tackle the problem from the other end by exploring what kind of moral machines are possible based on what computational systems can or cannot do. To do so, we analyze normative (...) ethics through the lens of computational complexity. First, we introduce computational complexity for the uninitiated reader and discuss how the complexity of ethical problems can be framed within Marr’s three levels of analysis. We then study a range of ethical problems based on consequentialism, deontology, and virtue ethics, with the aim of elucidating the complexity associated with the problems themselves (e.g., due to combinatorics, uncertainty, strategic dynamics), the computational methods employed (e.g., probability, logic, learning), and the available resources (e.g., time, knowledge, learning). The results indicate that most problems the normative frameworks pose lead to tractability issues in every category analyzed. Our investigation also provides several insights about the computational nature of normative ethics, including the differences between rule- and outcome-based moral strategies, and the implementation-variance with regard to moral resources. We then discuss the consequences complexity results have for the prospect of moral machines in virtue of the trade-off between optimality and efficiency. Finally, we elucidate how computational complexity can be used to inform both philosophical and cognitive-psychological research on human morality by advancing the moral tractability thesis. (shrink)

The same method that creates adversarial examples to fool image-classifiers can be used to generate counterfactual explanations that explain algorithmic decisions. This observation has led researchers to consider CEs as AEs by another name. We argue that the relationship to the true label and the tolerance with respect to proximity are two properties that formally distinguish CEs and AEs. Based on these arguments, we introduce CEs, AEs, and related concepts mathematically in a common framework. Furthermore, we show connections between current (...) methods for generating CEs and AEs, and estimate that the fields will merge more and more as the number of common use-cases grows. (shrink)

Technologies that deploy data science methods are liable to result in epistemic harms involving the diminution of individuals with respect to their standing as knowers or their credibility as sources of testimony. Not all harms of this kind are unjust but when they are we ought to try to prevent or correct them. Epistemically unjust harms will typically intersect with other more familiar and well-studied kinds of harm that result from the design, development, and use of data science technologies. However, (...) we argue that epistemic injustices can be distinguished conceptually from more familiar kinds of harm. We argue that epistemic harms are morally relevant even in cases where those who suffer them are unharmed in other ways. Via a series of examples from the criminal justice system, workplace hierarchies, and educational contexts we explain the kinds of epistemic injustice that can result from common uses of data science technologies. (shrink)

We show how the Full Bayesian Significance Test (FBST) can be used as a model selection criterion. The FBST was presented by Pereira and Stern as a coherent Bayesian significance test. Key Words: Bayesian test; Evidence; Global optimization; Information; Model selection; Numerical integration; Posterior density; Precise hypothesis; Regularization. AMS: 62A15; 62F15; 62H15.

This paper forges links between early analytic philosophy and the posits of semiotics. I show that there are some striking and potentially quite important, but perhaps unrecognized, connections between three key concepts in Wittgenstein’s middle and later philosophy, namely, complex, rule-following, and language games. This reveals the existence of a conceptual continuity between Wittgenstein’s “early” and “later” philosophy that can be applied to the analysis of the iterability of representation in computer-generated images. Methodologically, this paper clarifies to at least some (...) degree, the nature, progress and promise of an approach to doing philosophy and semiotics from a modally modest perspective that sees in the intellectual products of humanities, and not in unreflective empiricism, the future of scientific development. This hybrid, non-reductionist approach shows, among other things, that semiotic processes are encoded by specific types of complexes in computer-generated images that display iterability in time and space. (shrink)

Emotional design is an important development trend of interaction design. Emotional design in products plays a key role in enhancing user experience and inducing user emotional resonance. In recent years, based on the user's emotional experience, the design concept of strengthening product emotional design has become a new direction for most designers to improve their design thinking. In the emotional interaction design, the machine needs to capture the user's key information in real time, recognize the user's emotional state, and use (...) a variety of clues to finally determine the appropriate user model. Based on this background, this research uses a deep learning mechanism for more accurate and effective emotion recognition, thereby optimizing the design of the interactive system and improving the user experience. First of all, this research discusses how to use user characteristics such as speech, facial expression, video, heartbeat, etc., to make machines more accurately recognize human emotions. Through the analysis of various characteristics, the speech is selected as the experimental material. Second, a speech-based emotion recognition method is proposed. The mel-Frequency cepstral coefficient of the speech signal is used as the input of the improved long and short-term memory network. To ensure the integrity of the information and the accuracy of the output at the next moment, ILSTM makes peephole connections in the forget gate and input gate of LSTM, and adds the unit state as input data to the threshold layer. The emotional features obtained by ILSTM are input into the attention layer, and the self-attention mechanism is used to calculate the weight of each frame of speech signal. The speech features with higher weights are used to distinguish different emotions and complete the emotion recognition of the speech signal. Experiments on the EMO-DB and CASIA datasets verify the effectiveness of the model for emotion recognition. Finally, the feasibility of emotional interaction system design is discussed. (shrink)

We propose a formal framework for interpretable machine learning. Combining elements from statistical learning, causal interventionism, and decision theory, we design an idealised explanation game in which players collaborate to find the best explanation for a given algorithmic prediction. Through an iterative procedure of questions and answers, the players establish a three-dimensional Pareto frontier that describes the optimal trade-offs between explanatory accuracy, simplicity, and relevance. Multiple rounds are played at different levels of abstraction, allowing the players to explore overlapping causal (...) patterns of variable granularity and scope. We characterise the conditions under which such a game is almost surely guaranteed to converge on a optimal explanation surface in polynomial time, and highlight obstacles that will tend to prevent the players from advancing beyond certain explanatory thresholds. The game serves a descriptive and a normative function, establishing a conceptual space in which to analyse and compare existing proposals, as well as design new and improved solutions. (shrink)

The free energy principle says that any self-organising system that is at nonequilibrium steady-state with its environment must minimize its free energy. It is proposed as a grand unifying principle for cognitive science and biology. The principle can appear cryptic, esoteric, too ambitious, and unfalsifiable—suggesting it would be best to suspend any belief in the principle, and instead focus on individual, more concrete and falsifiable ‘process theories’ for particular biological processes and phenomena like perception, decision and action. Here, I explain (...) the free energy principle, and I argue that it is best understood as offering a conceptual and mathematical analysis of the concept of existence of self-organising systems. This analysis offers a new type of solution to long-standing problems in neurobiology, cognitive science, machine learning and philosophy concerning the possibility of normatively constrained, self-supervised learning and inference. The principle can therefore uniquely serve as a regulatory principle for process theories, to ensure that process theories conforming to it enable self-supervision. This is, at least for those who believe self-supervision is a foundational explanatory task, good reason to believe the free energy principle. (shrink)

We compare Karl Popper’s ideas concerning the falsifiability of a theory with similar notions from the part of statistical learning theory known as VC-theory . Popper’s notion of the dimension of a theory is contrasted with the apparently very similar VC-dimension. Having located some divergences, we discuss how best to view Popper’s work from the perspective of statistical learning theory, either as a precursor or as aiming to capture a different learning activity.

My dissertation explores the ways in which Rudolf Carnap sought to make philosophy scientific by further developing recent interpretive efforts to explain Carnap’s mature philosophical work as a form of engineering. It does this by looking in detail at his philosophical practice in his most sustained mature project, his work on pure and applied inductive logic. I, first, specify the sort of engineering Carnap is engaged in as involving an engineering design problem and then draw out the complications of design (...) problems from current work in history of engineering and technology studies. I then model Carnap’s practice based on those lessons and uncover ways in which Carnap’s technical work in inductive logic takes some of these lessons on board. This shows ways in which Carnap’s philosophical project subtly changes right through his late work on induction, providing an important corrective to interpretations that ignore the work on inductive logic. Specifically, I show that paying attention to the historical details of Carnap’s attempt to apply his work in inductive logic to decision theory and theoretical statistics in the 1950s and 1960s helps us understand how Carnap develops and rearticulates the philosophical point of the practical/theoretical distinction in his late work, offering thus a new interpretation of Carnap’s technical work within the broader context of philosophy of science and analytical philosophy in general. (shrink)

We describe reversible adaptive trees, a class of stochastic algorithms modified from the formerly described adaptive trees. They evolve in time a finite subset of an ambient Euclidean space of any dimension, starting from a seed point and, accreting points to the evolving set, they grow branches towards a target set which can depend on time. In contrast with plain adaptive trees, which were formerly proven to have strong convergence properties to a static target, the points of reversible adaptive trees (...) are removed from the tree when they have not been used recently enough in a path from the root to an accreted point. This, together with a straightening process performed on the branches, permits the tree to follow some moving targets and still remain adapted to it. We then discuss in what way one can see such reversible trees as a model for a qualitative property of resilience, which leads us to discuss qualitative modeling. (shrink)

The paper demonstrates that falsifiability is fundamental to learning. We prove the following theorem for statistical learning and sequential prediction: If a theory is falsifiable then it is learnable -- i.e. admits a strategy that predicts optimally. An analogous result is shown for universal induction.

Languages are transmitted from person to person and generation to generation via a process of iterated learning: people learn a language from other people who once learned that language themselves. We analyze the consequences of iterated learning for learning algorithms based on the principles of Bayesian inference, assuming that learners compute a posterior distribution over languages by combining a prior (representing their inductive biases) with the evidence provided by linguistic data. We show that when learners sample languages from this posterior (...) distribution, iterated learning converges to a distribution over languages that is determined entirely by the prior. Under these conditions, iterated learning is a form of Gibbs sampling, a widely-used Markov chain Monte Carlo algorithm. The consequences of iterated learning are more complicated when learners choose the language with maximum posterior probability, being affected by both the prior of the learners and the amount of information transmitted between generations. We show that in this case, iterated learning corresponds to another statistical inference algorithm, a variant of the expectation-maximization (EM) algorithm. These results clarify the role of iterated learning in explanations of linguistic universals and provide a formal connection between constraints on language acquisition and the languages that come to be spoken, suggesting that information transmitted via iterated learning will ultimately come to mirror the minds of the learners. (shrink)

Vector-based models of word meaning have become increasingly popular in cognitive science. The appeal of these models lies in their ability to represent meaning simply by using distributional information under the assumption that words occurring within similar contexts are semantically similar. Despite their widespread use, vector-based models are typically directed at representing words in isolation, and methods for constructing representations for phrases or sentences have received little attention in the literature. This is in marked contrast to experimental evidence (e.g., in (...) sentential priming) suggesting that semantic similarity is more complex than simply a relation between isolated words. This article proposes a framework for representing the meaning of word combinations in vector space. Central to our approach is vector composition, which we operationalize in terms of additive and multiplicative functions. Under this framework, we introduce a wide range of composition models that we evaluate empirically on a phrase similarity task. (shrink)

The idea that “simplicity is a sign of truth”, and the related “Occam’s razor” principle, stating that, all other things being equal, simpler models should be preferred to more complex ones, have been long discussed in philosophy and science. We explore these ideas in the context of supervised machine learning, namely the branch of artificial intelligence that studies algorithms which balance simplicity and accuracy in order to effectively learn about the features of the underlying domain. Focusing on statistical learning theory, (...) we show that situations exist for which a preference for simpler models provably slows down, instead of favoring, the supervised learning process. Our results shed new light on the relations between simplicity and truth approximation, which are briefly discussed in the context of both machine learning and the philosophy of science. (shrink)

Rhetorical Structure Theory is a theory of text organization that has led to areas of application beyond discourse analysis and text generation, its original goals. In this article, we review the most important applications in several areas: discourse analysis, theoretical linguistics, psycholinguistics, and computational linguistics. We also provide a list of resources useful for work within the RST framework. The present article is a complement to our review of the theoretical aspects of the theory.

As we increasingly delegate decision-making to algorithms, whether directly or indirectly, important questions emerge in circumstances where those decisions have direct consequences for individual rights and personal opportunities, as well as for the collective good. A key problem for policymakers is that the social implications of these new methods can only be grasped if there is an adequate comprehension of their general technical underpinnings. The discussion here focuses primarily on the case of enforcement decisions in the criminal justice system, but (...) draws on similar situations emerging from other algorithms utilised in controlling access to opportunities, to explain how machine learning works and, as a result, how decisions are made by modern intelligent algorithms or 'classifiers'. It examines the key aspects of the performance of classifiers, including how classifiers learn, the fact that they operate on the basis of correlation rather than causation, and that the term 'bias' in machine learning has a different meaning to common usage. An example of a real world 'classifier', the Harm Assessment Risk Tool (HART), is examined, through identification of its technical features: the classification method, the training data and the test data, the features and the labels, validation and performance measures. Four normative benchmarks are then considered by reference to HART: (a) prediction accuracy (b) fairness and equality before the law (c) transparency and accountability (d) informational privacy and freedom of expression, in order to demonstrate how its technical features have important normative dimensions that bear directly on the extent to which the system can be regarded as a viable and legitimate support for, or even alternative to, existing human decision-makers. (shrink)

The present paper shows how statistical learning theory and machine learning models can be used to enhance understanding of AI-related epistemological issues regarding inductive reasoning and reliability of generalisations. Towards this aim, the paper proceeds as follows. First, it expounds Price’s dual image of representation in terms of the notions of e-representations and i-representations that constitute subject naturalism. For Price, this is not a strictly anti-representationalist position but rather a dualist one (e- and i-representations). Second, the paper links this debate (...) with machine learning in terms of statistical learning theory becoming more viable epistemological tool when it abandons the perspective of object naturalism. The paper then argues that machine learning grounds a form of knowing that can be understood in terms of e- and i-representation learning. Third, this synthesis shows a way of analysing inductive reasoning in terms of reliability of generalisations stemming from a structure of e- and i-representations. In the age of Artificial Intelligence, connecting Price’s dual view of representation with Deep Learning provides an epistemological way forward and even perhaps an approach to how knowing is possible. (shrink)

Nelson Goodman's new riddle of induction forcefully illustrates a challenge that must be confronted by any adequate theory of inductive inference: provide some basis for choosing among alternative hypotheses that fit past data but make divergent predictions. One response to this challenge is to distinguish among alternatives by means of some epistemically significant characteristic beyond fit with the data. Statistical learning theory takes this approach by showing how a concept similar to Popper's notion of degrees of testability is linked to (...) minimizing expected predictive error. In contrast, formal learning theory appeals to Ockham's razor, which it justifies by reference to the goal of enhancing efficient convergence to the truth. In this essay, I show that, despite their differences, statistical and formal learning theory yield precisely the same result for a class of inductive problems that I call strongly VC ordered, of which Goodman's riddle is just one example. (shrink)

Recently, e-learning systems for self-learning with various types of retrieval functions have been developed. This paper describes a method for keyword extraction using retrieval information stored from many students through the retrieval functions. Firstly, we show that teachers tend to consider technical terms as important, while students unfamiliar with the technical terms tend to retrieve the terms, therefore there is a clear correlation between keywords extracted by the teachers and the retrieval words by the students. Secondly, we propse a method (...) utilizing retrieval information from the students for keyword extraction, and show that the mehod can achieve quite better performance than a method extracting keywords using only lecture information. (shrink)

This paper presents a computational model of the way humans inductively identify and aggregate concepts from the low-level stimuli they are exposed to. Based on the idea that humans tend to select the simplest structures, it implements a dynamic hierarchical chunking mechanism in which the decision whether to create a new chunk is based on an information-theoretic criterion, the Minimum Description Length (MDL) principle. We present theoretical justifications for this approach together with results of an experiment in which participants, exposed (...) to meaningless symbols, have been implicitly encouraged to create high-level concepts by grouping them. Results show that the designed model, called hereafter MDLChunker, makes precise quantitative predictions both on the kind of chunks created by the participants and also on the moment at which these creations occur. They suggest that the simplicity principle used to design MDLChunker is particularly efficient to model chunking mechanisms. The main interest of this model over existing ones is that it does not require any adjustable parameter. (shrink)

The import of computational learning theories and techniques on the ethics of human-robot interaction is explored in the context of recent developments of personal robotics. An epistemological reflection enables one to isolate a variety of background hypotheses that are needed to achieve successful learning from experience in autonomous personal robots. The conjectural character of these background hypotheses brings out theoretical and practical limitations in our ability to predict and control the behaviour of learning robots in their interactions with humans. Responsibility (...) ascription problems, which concern damages caused by learning robot actions, are analyzed in the light of these epistemic limitations. Finally, a broad framework is outlined for ethically motivated scientific inquiries, which aim at improving our capability to understand, anticipate, and selectively cope with harmful errors by learning robots. (shrink)

The philosophical innateness debate has long relied onpsychological evidence. For a century, however, a parallel debate hastaken place within neuroscience. In this paper, I consider theimplications of this neuroscience debate for the philosophicalinnateness debate. By combining the tools of theoretical neurobiologyand learning theory, I introduce the ``problem of development'' that alladaptive systems must solve, and suggest how responses to this problemcan demarcate a number of innateness proposals. From this perspective, Isuggest that the majority of natural systems are in fact innate. (...) Lastly,I consider the acquistion strategies implemented by the human brain andsuggest that there is a rigorous way of characterizing these ``neuralconstructivist'' strategies as not being strongly innate. Alternatives toinnateness are thus both rigorously definable and empirically supported. (shrink)

Cancer is fast becoming an alarming cause of human death. However, it has been reported that if the disease is detected at an early stage, diagnosed, treated appropriately, the patient has better chances of survival long life. Machine learning technique with feature-selection contributes greatly to the detecting of cancer, because an efficient feature-selection method can remove redundant features. In this paper, a Fuzzy Preference-Based Rough Set (FPRS) blended with Support Vector Machine (SVM) has been applied in order to predict cancer (...) biomarkers for biological and gene expression datasets. Biomarkers are determined by deploying three models of FPRS, namely, Fuzzy Upward Consistency (FUC), Fuzzy Downward Consistency (FLC), and Fuzzy Global Consistency (FGC). The efficiency of the three models with SVM on five datasets is exhibited, and the biomarkers that have been identified from FUC models have been reported. (shrink)

Our best sciences are frequently held to be one way, perhaps the optimal way, to learn about the world’s higher-level ontology and structure. I first argue that which scientific theory is “best” depends in part on our goals or purposes. As a result, it is theoretically possible to have two scientific theories of the same domain, where each theory is best for some goal, but where the two theories posit incompatible ontologies. That is, it is possible for us to have (...) goal-dependent pluralism in our scientific ontologies. This ontological pluralism arises simply from our inability to directly know the world’s objects, rather than any particular claims about our cognitive limits, values, or social structures. I then present two case studies in which this possibility actually occurs—one based on simulations and theoretical analyses of constructed causal systems, and one from actual scientific investigations into the proper ontology for ocean regions. (shrink)

Bayesian probability is normally defined over a fixed language or eventspace. But in practice language is susceptible to change, and thequestion naturally arises as to how Bayesian degrees of belief shouldchange as language changes. I argue here that this question poses aserious challenge to Bayesianism. The Bayesian may be able to meet thischallenge however, and I outline a practical method for changing degreesof belief over changes in finite propositional languages.

Real engines of the artificial intelligence revolution, machine learning models, and algorithms are embedded nowadays in many services and products around us. As a society, we argue it is now necessary to transition into a phronetic paradigm focused on the ethical dilemmas stemming from the conception and application of AIs to define actionable recommendations as well as normative solutions. However, both academic research and society-driven initiatives are still quite far from clearly defining a solid program of study and intervention. In (...) this contribution, we will focus on selected ethical investigations around AI by proposing an incremental model of trust that can be applied to both human-human and human-AI interactions. Starting with a quick overview of the existing accounts of trust, with special attention to Taddeo’s concept of “e-trust,” we will discuss all the components of the proposed model and the reasons to trust in human-AI interactions in an example of relevance for business organizations. We end this contribution with an analysis of the epistemic and pragmatic reasons of trust in human-AI interactions and with a discussion of kinds of normativity in trustworthiness of AIs. (shrink)

Both dilation and non-conglomerability have been alleged to conflict with a fundamental principle of Bayesian methodology that we call \textit{Good's Principle}: one should always delay making a terminal decision between alternative courses of action if given the opportunity to first learn, at zero cost, the outcome of an experiment relevant to the decision. In particular, both dilation and non-conglomerability have been alleged to permit or even mandate choosing to make a terminal decision in deliberate ignorance of relevant, cost-free information. Although (...) dilation and non-conglomerability share some similarities, some authors maintain that there are important differences between the two that warrant endorsing different normative positions regarding dilation and non-conglomerability. This article reassesses the grounds for treating dilation and non-conglomerability differently. Our analysis exploits a new and general characterization result for dilation to draw a closer connection between dilation and non-conglomerability. (shrink)

We introduce a new convolution kernel for labeled ordered trees with arbitrary subgraph features, and an efficient algorithm for computing the kernel with the same time complexity as that of the parse tree kernel. The proposed kernel is extended to allow mutations of labels and structures without increasing the order of computation time. Moreover, as a limit of generalization of the tree kernels, we show a hardness result in computing kernels for unordered rooted labeled trees with arbitrary subgraph features.

Brain Computer Interfaces (BCIs) enable one to control peripheral ICT and robotic devices by processing brain activity on-line. The potential usefulness of BCI systems, initially demonstrated in rehabilitation medicine, is now being explored in education, entertainment, intensive workflow monitoring, security, and training. Ethical issues arising in connection with these investigations are triaged taking into account technological imminence and pervasiveness of BCI technologies. By focussing on imminent technological developments, ethical reflection is informatively grounded into realistic protocols of brain-to-computer communication. In particular, (...) it is argued that human-machine adaptation and shared control distinctively shape autonomy and responsibility issues in current BCI interaction environments. Novel personhood issues are identified and analyzed too. These notably concern (i) the “sub-personal” use of human beings in BCI-enabled cooperative problem solving, and (ii) the pro-active protection of personal identity which BCI rehabilitation therapies may afford, in the light of so-called motor theories of thinking, for the benefit of patients affected by severe motor disabilities. (shrink)

Within the last decade, several approaches using quaternion numbers to handle and model multiband images in a holistic manner were introduced. The quaternion Fourier transform can be efficiently used to model texture in multidimensional data such as color images. For practical application, multispectral satellite data appear as a primary source for measuring past trends and monitoring changes in forest carbon stocks. In this work, we propose a texture-color descriptor based on the quaternion Fourier transform to extract relevant information from multiband (...) satellite images. We propose a new multiband image texture model extraction, called FOTO++, in order to address biomass estimation issues. The first stage consists in removing noise from the multispectral data while preserving the edges of canopies. Afterward, color texture descriptors are extracted thanks to a discrete form of the quaternion Fourier transform, and finally the support vector regression method is used to deduce biomass estimation from texture indices. Our texture features are modeled using a vector composed with the radial spectrum coming from the amplitude of the quaternion Fourier transform. We conduct several experiments in order to study the sensitivity of our model to acquisition parameters. We also assess its performance both on synthetic images and on real multispectral images of Cameroonian forest. The results show that our model is more robust to acquisition parameters than the classical Fourier Texture Ordination model. Our scheme is also more accurate for aboveground biomass estimation. We stress that a similar methodology could be implemented using quaternion wavelets. These results highlight the potential of the quaternion-based approach to study multispectral satellite images. (shrink)