In this commentary, I advocate for strict relations between Marr's levels of analysis. Under a strict relationship, each level is exactly implemented by the subordinate level. This yields two benefits. First, it brings consistency for multilevel explanations. Second, similar to how a sculptor chisels away superfluous marble, a modeler can chisel a computational-level model by applying constraints. By sculpting the model, one restricts the set of possible algorithmic- and implementational-level theories.

I extend Jones & Love's (J&L's) critique of Bayesian models and evaluate the conceptual foundations on which they are built. I argue that: (1) the part of Bayesian models is scientifically trivial; (2) theory is a fiction that arises from an inappropriate programming metaphor; and (3) the real scientific problems lie outside Bayesian theorizing.

According to Marr, a computational-level theory consists of two elements, the what and the why . This article highlights the distinct role of the Why element in the computational analysis of vision. Three theses are advanced: ( a ) that the Why element plays an explanatory role in computational-level theories, ( b ) that its goal is to explain why the computed function (specified by the What element) is appropriate for a given visual task, (...) and ( c ) that the explanation consists in showing that the functional relations between the representing cells are similar to the “external” mathematical relations between the entities that these cells represent. *Received September 2009; revised January 2010. †To contact the author, please write to: Departments of Philosophy and Cognitive Science, The Hebrew University, Jerusalem 91905, Israel; e-mail: [email protected]. (shrink)

According to Marr, a computational-level theory consists of two elements, the what and the why. This article highlights the distinct role of the Why element in the computational analysis of vision. Three theses are advanced: that the Why element plays an explanatory role in computational-level theories, that its goal is to explain why the computed function is appropriate for a given visual task, and that the explanation consists in showing that the functional relations between (...) the representing cells are similar to the “external” mathematical relations between the entities that these cells represent. (shrink)

Leech et al. present a connectionist algorithm as a model of (the development) of analogizing, but they do not specify the algorithm's associated computational-level theory, nor its computational complexity. We argue that doing so may be essential for connectionist cognitive models to have full explanatory power and transparency, as well as for assessing their scalability to real-world input domains.

We sketch four applications of Marr's levels-of-analysis methodology to the relations between logic and experimental data in the cognitive neuroscience of language and reasoning. The first part of the paper illustrates the explanatory power of computational level theories based on logic. We show that a Bayesian treatment of the suppression task in reasoning with conditionals is ruled out by EEG data, supporting instead an analysis based on defeasible logic. Further, we describe how results from an EEG (...) study on temporal prepositions can be reanalyzed using formal semantics, addressing a potential confound. The second part of the article demonstrates the predictive power of logical theories drawing on EEG data on processing progressive constructions and on behavioral data on conditional reasoning in people with autism. Logical theories can constrain processing hypotheses all the way down to neurophysiology, and conversely neuroscience data can guide the selection of alternative computational level models of cognition. (shrink)

The general problem of visual search can be shown to be computationally intractable in a formal, complexity-theoretic sense, yet visual search is extensively involved in everyday perception, and biological systems manage to perform it remarkably well. Complexity level analysis may resolve this contradiction. Visual search can be reshaped into tractability through approximations and by optimizing the resources devoted to visual processing. Architectural constraints can be derived using the minimum cost principle to rule out a large class of potential (...) solutions. The evidence speaks strongly against bottom-up approaches to vision. In particular, the constraints suggest an attentional mechanism that exploits knowledge of the specific problem being solved. This analysis of visual search performance in terms of attentional influences on visual information processing and complexity satisfaction allows a large body of neurophysiological and psychological evidence to be tied together. (shrink)

We sketch four applications of Marr's levels‐of‐analysis methodology to the relations between logic and experimental data in the cognitive neuroscience of language and reasoning. The first part of the paper illustrates the explanatory power of computational level theories based on logic. We show that a Bayesian treatment of the suppression task in reasoning with conditionals is ruled out by EEG data, supporting instead an analysis based on defeasible logic. Further, we describe how results from an EEG (...) study on temporal prepositions can be reanalyzed using formal semantics, addressing a potential confound. The second part of the article demonstrates the predictive power of logical theories drawing on EEG data on processing progressive constructions and on behavioral data on conditional reasoning in people with autism. Logical theories can constrain processing hypotheses all the way down to neurophysiology, and conversely neuroscience data can guide the selection of alternative computational level models of cognition. (shrink)

Marr's levels of analysis—computational, algorithmic, and implementation—have served cognitive science well over the last 30 years. But the recent increase in the popularity of the computational level raises a new challenge: How do we begin to relate models at different levels of analysis? We propose that it is possible to define levels of analysis that lie between the computational and the algorithmic, providing a way to build a bridge between computational- and algorithmic- (...) class='Hi'>level models. The key idea is to push the notion of rationality, often used in defining computational-level models, deeper toward the algorithmic level. We offer a simple recipe for reverse-engineering the mind's cognitive strategies by deriving optimal algorithms for a series of increasingly more realistic abstract computational architectures, which we call “resource-rational analysis.”. (shrink)

The literature on self-driving cars and ethics continues to grow. Yet much of it focuses on ethical complexities emerging from an individual vehicle. That is an important but insufficient step towards determining how the technology will impact human lives and society more generally. What must complement ongoing discussions is a broader, system level of analysis that engages with the interactions and effects that these cars will have on one another and on the socio-technical systems in which they are (...) embedded. To bring the conversation of self-driving cars to the system level, we make use of two traffic scenarios which highlight some of the complexities that designers, policymakers, and others should consider related to the technology. We then describe three approaches that could be used to address such complexities and their associated shortcomings. We conclude by bringing attention to the “Moral Responsibility for Computing Artifacts: The Rules”, a framework that can provide insight into how to approach ethical issues related to self-driving cars. (shrink)

Every scientist chooses a preferred level of analysis and this choice shapes the research program, even determining what counts as evidence. This contribution revisits Marr's three levels of analysis and evaluates the prospect of making progress at each individual level. After reviewing limitations of theorizing within a level, two strategies for integration across levels are considered. One is top–down in that it attempts to build a bridge from the computational to algorithmic level. Limitations (...) of this approach include insufficient theoretical constraint at the computation level to provide a foundation for integration, and that people are suboptimal for reasons other than capacity limitations. Instead, an inside-out approach is forwarded in which all three levels of analysis are integrated via the algorithmic level. This approach maximally leverages mutual data constraints at all levels. For example, algorithmic models can be used to interpret brain imaging data, and brain imaging data can be used to select among competing models. Examples of this approach to integration are provided. This merging of levels raises questions about the relevance of Marr's tripartite view. (shrink)

In its origins educational neuroscience has started as an endeavor to discuss implications of neuroscience studies for education. However, it is now on its way to become a transdisciplinary field, incorporating findings, theoretical frameworks and methodologies from education, and cognitive and brain sciences. Given the differences and diversity in the originating disciplines, it has been a challenge for educational neuroscience to integrate both theoretical and methodological perspective in education and neuroscience in a coherent way. We present a multi-level framework (...) for educational neuroscience, which argues for integration of multiple levels of analysis, some originating in brain and cognitive sciences, others in education, as a roadmap for the future of educational neuroscience with concrete examples in moral education. (shrink)

We show how faceted search using a combination of traditional classification systems and mixed-membership topic models can go beyond keyword search to inform resource discovery, hypothesis formulation, and argument extraction for interdisciplinary research. Our test domain is the history and philosophy of scientific work on animal mind and cognition. The methods can be generalized to other research areas and ultimately support a system for semi-automatic identification of argument structures. We provide a case study for the application of the methods to (...) the problem of identifying and extracting arguments about anthropomorphism during a critical period in the development of comparative psychology. We show how a combination of classification systems and mixed-membership models trained over large digital libraries can inform resource discovery in this domain. Through a novel approach of “drill-down” topic modeling—simultaneously reducing both the size of the corpus and the unit of analysis—we are able to reduce a large collection of fulltext volumes to a much smaller set of pages within six focal volumes containing arguments of interest to historians and philosophers of comparative psychology. The volumes identified in this way did not appear among the first ten results of the keyword search in the HathiTrust digital library and the pages bear the kind of “close reading” needed to generate original interpretations that is the heart of scholarly work in the humanities. Zooming back out, we provide a way to place the books onto a map of science originally constructed from very different data and for different purposes. The multilevel approach advances understanding of the intellectual and societal contexts in which writings are interpreted. (shrink)

Computer simulations can be useful tools to support philosophers in validating their theories, especially when these theories concern phenomena showing nontrivial dynamics. Such theories are usually informal, whilst for computer simulation a formally described model is needed. In this paper, a methodology is proposed to gradually formalise philosophical theories in terms of logically formalised dynamic properties. One outcome of this process is an executable logic-based temporal specification, which within a dedicated software environment can be used as a simulation model to (...) perform simulations. This specification provides a logical formalisation at the lowest aggregation level of the basic mechanisms underlying a process. In addition, dynamic properties at a higher aggregation level that may emerge from the mechanisms specified by the lower level properties, can be specified. Software tools are available to support specification, and to automatically check such higher level properties against the lower level properties and against generated simulation traces. As an illustration, three case studies are discussed showing successful applications of the approach to formalise and analyse, among others, Clark’s theory on extended mind, Damasio’s theory on core consciousness, and Dennett’s perspective on intertemporal decision making and altruism. (shrink)

Algorithmic-level specifications carry part of the explanatory burden in most psychological theories. It is, thus, inappropriate to limit a comparison and evaluation of theories to the computational level. A rational analysis considers people's goal-directed and environmentally adaptive rationality; it is not normative. Adaptive rationality is by definition non-absolute; hence, neither deductive logic nor Bayesian probability theory has absolute normative status.

Herpes simplex virus type 2 is the most prevalent sexually transmitted infection in the world, despite the availability of effective anti-viral treatments. A mathematical model to explore the association between gender and HSV-2 treatment adherence is developed. Threshold parameters are determined and stabilities analyzed. Sensitivity analysis of the reproduction number and the numerical simulations suggest that treatment adherence for both females and males are equally important in keeping the reproduction as low as possible. The basic model is then extended (...) to incorporate time-dependent intervention strategies. The Pontryagin’s Maximum Principle is used to characterize the optimal level of the controls, and the resulting optimality system is solved numerically. (shrink)

Are all three of Marr's levels needed? Should they be kept distinct? We argue for the distinct contributions and methodologies of each level of analysis. It is important to maintain them because they provide three different perspectives required to understand mechanisms, especially information-processing mechanisms. The computational perspective provides an understanding of how a mechanism functions in broader environments that determines the computations it needs to perform. The representation and algorithmic perspective offers an understanding of how information about (...) the environment is encoded within the mechanism and what are the patterns of organization that enable the parts of the mechanism to produce the phenomenon. The implementation perspective yields an understanding of the neural details of the mechanism and how they constrain function and algorithms. Once we adequately characterize the distinct role of each level of analysis, it is fairly straightforward to see how they relate. (shrink)

This article presents the results of an empirical study of the “sportification” of amateur-level competitive computer gaming. How do amateur players, who are unlikely to become professional esports players, turn what is considered to be enjoyable entertainment into a collective activity that demonstrates traits traditionally associated with professional sports, such as self-discipline, a focus on achieving results and overcoming personal limitations? Ethnographic research, consisting of in-depth interviews and participant observations, was conducted in the last quarter of 2019 based on (...) a series of tournaments (Rocket League, Counter-Strike: Global Offensive, League of Legends, and Rainbow Six Siege) managed by a student esports club of a Moscow-based university. Such clubs combine different levels of esports practice: players who consider themselves amateurs who attend tournaments “for fun”, players seeking professionalization and who have already received the first modest revenue from their activity, as well as players who achieved the status of a cyberathlete through membership in one of the university teams. Thus, student esports represents an ideal environment for exploring the dynamics of changes in the meanings that players put into their activity, as well as for finding elements holding to­gether what could be called “esports culture”. The analysis conducted in the present article is based not so much on the interpretations of the phe­nomenon of esports that exists in the field of game studies, as on two clas­sical sociological traditions: the writings of Erving Goffman and Howard Becker, on the one hand, and those of Max Weber and Norbert Elias on the other. As a result, the process of the “sportification” of the amateur practice of competitive computer games is presented in the article as a change in the actors’ level of reflexivity: towards the game, towards themselves during play, and towards the world around the game, occurring as they move through the stages of the “career of a competitive player”. ­. (shrink)

Computational models of child language development can help us understand the cognitive underpinnings of the language learning process, which occurs along several linguistic levels at once (e.g., prosodic and phonological). However, in light of the replication crisis, modelers face the challenge of selecting representative and consolidated infant data. Thus, it is desirable to have evaluation methodologies that could account for robust empirical reference data, across multiple infant capabilities. Moreover, there is a need for practices that can compare developmental trajectories (...) of infants to those of models as a function of language experience and development. The present study aims to take concrete steps to address these needs by introducing the concept of comparing models with large-scale cumulative empirical data from infants, as quantified by meta-analyses conducted across a large number of individual behavioral studies. We formalize the connection between measurable model and human behavior, and then present a conceptual framework for meta-analytic evaluation of computational models. We exemplify the meta-analytic model evaluation approach with two modeling experiments on infant-directed speech preference and native/non-native vowel discrimination. (shrink)

Thirty years after the publication of Marr's seminal book Vision the papers in this topic consider the contemporary status of his influential conception of three distinct levels of analysis for information-processing systems, and in particular the role of the algorithmic and representational level with its cognitive-level concepts. This level has been downplayed or eliminated both by reductionist neuroscience approaches from below that seek to account for behavior from the implementation level and by Bayesian approaches from (...) above that seek to account for behavior in purely computational-level terms. (shrink)

The computational view of mind rests on certain intuitions regarding the fundamental similarity between computation and cognition. We examine some of these intuitions and suggest that they derive from the fact that computers and human organisms are both physical systems whose behavior is correctly described as being governed by rules acting on symbolic representations. Some of the implications of this view are discussed. It is suggested that a fundamental hypothesis of this approach is that there is a natural domain (...) of human functioning that can be addressed exclusively in terms of a formal symbolic or algorithmic vocabulary or level of analysis. Much of the paper elaborates various conditions that need to be met if a literal view of mental activity as computation is to serve as the basis for explanatory theories. The coherence of such a view depends on there being a principled distinction between functions whose explanation requires that we posit internal representations and those that we can appropriately describe as merely instantiating causal physical or biological laws. In this paper the distinction is empirically grounded in a methodological criterion called the " cognitive impenetrability condition." Functions are said to be cognitively impenetrable if they cannot be influenced by such purely cognitive factors as goals, beliefs, inferences, tacit knowledge, and so on. Such a criterion makes it possible to empirically separate the fixed capacities of mind from the particular representations and algorithms used on specific occasions. In order for computational theories to avoid being ad hoc, they must deal effectively with the "degrees of freedom" problem by constraining the extent to which they can be arbitrarily adjusted post hoc to fit some particular set of observations. This in turn requires that the fixed architectural function and the algorithms be independently validated. It is argued that the architectural assumptions implicit in many contemporary models run afoul of the cognitive impenetrability condition, since the required fixed functions are demonstrably sensitive to tacit knowledge and goals. The paper concludes with some tactical suggestions for the development of computational cognitive theories. (shrink)

This thesis is devoted to the exploration of the complexity of some mathematical problems using the framework of computable analysis and descriptive set theory. We will especially focus on Weihrauch reducibility as a means to compare the uniform computational strength of problems. After a short introduction of the relevant background notions, we investigate the uniform computational content of problems arising from theorems that lie at the higher levels of the reverse mathematics hierarchy.We first analyze the strength of (...) the open and clopen Ramsey theorems. Since there is not a canonical way to phrase these theorems as multi-valued functions, we identify eight different multi-valued functions and study their degree from the point of view of Weihrauch, strong Weihrauch, and arithmetic Weihrauch reducibility.We then discuss some new operators on multi-valued functions and study their algebraic properties and the relations with other previously studied operators on problems. In particular, we study the first-order part and the deterministic part of a problem f, capturing the Weihrauch degree of the strongest multi-valued problem that is reducible to f and that, respectively, has codomain $\mathbb {N}$ or is single-valued.These notions proved to be extremely useful when exploring the Weihrauch degree of the problem $\mathsf {DS}$ of computing descending sequences in ill-founded linear orders. They allow us to show that $\mathsf {DS}$, and the Weihrauch equivalent problem $\mathsf {BS}$ of finding bad sequences through non-well quasi-orders, while being very “hard” to solve, are rather weak in terms of uniform computational strength. We then generalize $\mathsf {DS}$ and $\mathsf {BS}$ by considering $\boldsymbol {\Gamma }$ -presented orders, where $\boldsymbol {\Gamma }$ is a Borel pointclass or $\boldsymbol {\Delta }^1_1$, $\boldsymbol {\Sigma }^1_1$, $\boldsymbol {\Pi }^1_1$. We study the obtained $\mathsf {DS}$ -hierarchy and $\mathsf {BS}$ -hierarchy of problems in comparison with the Baire hierarchy and show that they do not collapse at any finite level.Finally, we work in the context of geometric measure theory and we focus on the characterization, from the point of view of descriptive set theory, of some conditions involving the notions of Hausdorff/Fourier dimension and Salem sets. We first work in the hyperspace $\mathbf {K}$ of compact subsets of $[0,1]$ and show that the closed Salem sets form a $\boldsymbol {\Pi }^0_3$ -complete family. This is done by characterizing the complexity of the family of sets having sufficiently large Hausdorff or Fourier dimension. We also show that the complexity does not change if we increase the dimension of the ambient space and work in $\mathbf {K}$. We also generalize the results by relaxing the compactness of the ambient space and show that the closed Salem sets are still $\boldsymbol {\Pi }^0_3$ -complete when we endow $\mathbf {F}$ with the Fell topology. A similar result holds also for the Vietoris topology.We conclude by analyzing the same notions from the point of view of effective descriptive set theory and Type-2 Theory of Effectivity, and show that the complexities remain the same also in the lightface case. In particular, we show that the family of all the closed Salem sets is $\Pi ^0_3$ -complete. We furthermore characterize the Weihrauch degree of the functions computing Hausdorff and Fourier dimension of closed sets.prepared by Manlio Valenti.E-mail:[email protected]. (shrink)

One of the most lasting contributions of Marr's posthumous book is his articulation of the different “levels of analysis” that are needed to understand vision. Although a variety of work has examined how these different levels are related, there is comparatively little examination of the assumptions on which his proposed levels rest, or the plausibility of the approach Marr articulated given those assumptions. Marr placed particular significance on computational level theory, which specifies the “goal” of a computation, (...) its appropriateness for solving a particular problem, and the logic by which it can be carried out. The structure of computational level theory is inherently teleological: What the brain does is described in terms of its purpose. I argue that computational level theory, and the reverse-engineering approach it inspires, requires understanding the historical trajectory that gave rise to functional capacities that can be meaningfully attributed with some sense of purpose or goal, that is, a reconstruction of the fitness function on which natural selection acted in shaping our visual abilities. I argue that this reconstruction is required to distinguish abilities shaped by natural selection—“natural tasks” —from evolutionary “by-products”, rather than merely demonstrating that computational goals can be embedded in a Bayesian model that renders a particular behavior or process rational. (shrink)

In this study, we use temporally aligned word embeddings and a large diachronic corpus of English to quantify language change in a data-driven, scalable way, which is grounded in language use. We show a unique and reliable relation between measures of language change and age of acquisition (AoA) while controlling for frequency, contextual diversity, concreteness, length, dominant part of speech, orthographic neighborhood density, and diachronic frequency variation. We analyze measures of language change tackling both the change in lexical representations and (...) the change in the relation between lexical representations and the words with the most similar usage patterns, showing that they capture different aspects of language change. Our results show a unique relation between language change and AoA, which is stronger when considering neighborhood-level measures of language change: Words with more coherent diachronic usage patterns tend to be acquired earlier. The results support theories positing a link between ontogenetic and ethnogenetic processes in language. (shrink)

Whereas computer simulations involve no direct physical interaction between the machine they are run on and the physical systems they are used to investigate, they are often used as experiments and yield data about these systems. It is commonly argued that they do so because they are implemented on physical machines. We claim that physicality is not necessary for their representational and predictive capacities and that the explanation of why computer simulations generate desired information about their target system is only (...) to be found in the detailed analysis of their semantic levels. We provide such an analysis and we determine the actual consequences of physical implementation for simulations. (shrink)

Marr’s seminal distinction between computational, algorithmic, and implementational levels of analysis has inspired research in cognitive science for more than 30 years. According to a widely-used paradigm, the modelling of cognitive processes should mainly operate on the computational level and be targeted at the idealised competence, rather than the actual performance of cognisers in a specific domain. In this paper, we explore how this paradigm can be adopted and revised to understand mathematical problem solving. The (...) class='Hi'>computational-level approach applies methods from computational complexity theory and focuses on optimal strategies for completing cognitive tasks. However, human cognitive capacities in mathematical problem solving are essentially characterised by processes that are computationally sub-optimal, because they initially add to the computational complexity of the solutions. Yet, these solutions can be optimal for human cognisers given the acquisition and enactment of mathematical practices. Here we present diagrams and the spatial manipulation of symbols as two examples of problem solving strategies that can be computationally sub-optimal but humanly optimal. These aspects need to be taken into account when analysing competence in mathematical problem solving. Empirically informed considerations on enculturation can help identify, explore, and model the cognitive processes involved in problem solving tasks. The enculturation account of mathematical problem solving strongly suggests that computational-level analyses need to be complemented by considerations on the algorithmic and implementational levels. The emerging research strategy can help develop algorithms that model what we call enculturated cognitive optimality in an empirically plausible and ecologically valid way. (shrink)

The commentaries reflect three core themes that pertain not just to our theory, but to the enterprise of connectionist modeling more generally. The first concerns the relationship between a cognitive theory and an implemented computer model. Specifically, how does one determine, when a model departs from the theory it exemplifies, whether the departure is a useful simplification or a critical flaw? We argue that the answer to this question depends partially upon the model's intended function, and we suggest that connectionist (...) models have important functions beyond the commonly accepted goals of fitting data and making predictions. The second theme concerns perceived in-principle limitations of the connectionist approach to cognition, and the specific concerns these perceived limitations raise for our theory. We argue that the approach is not in fact limited in the ways our critics suggest. One common misconception, that connectionist models cannot address abstract or relational structure, is corrected through new simulations showing directly that such structure can be captured. The third theme concerns the relationship between parallel distributed processing (PDP) models and structured probabilistic approaches. In this case we argue that there the difference between the approaches is not merely one of levels. Our PDP approach differs from structured statistical approaches at all of Marr's levels, including the characterization of the goals of cognitive computations, and of the representations and algorithms used. (shrink)

We propose a framework for including information‐processing bounds in rational analyses. It is an application of bounded optimality (Russell & Subramanian, 1995) to the challenges of developing theories of mechanism and behavior. The framework is based on the idea that behaviors are generated by cognitive mechanisms that are adapted to the structure of not only the environment but also the mind and brain itself. We call the framework computational rationality to emphasize the incorporation of computational mechanism into the (...) definition of rational action. Theories are specified as optimal program problems, defined by an adaptation environment, a bounded machine, and a utility function. Such theories yield different classes of explanation, depending on the extent to which they emphasize adaptation to bounds, and adaptation to some ecology that differs from the immediate local environment. We illustrate this variation with examples from three domains: visual attention in a linguistic task, manual response ordering, and reasoning. We explore the relation of this framework to existing “levels” approaches to explanation, and to other optimality‐based modeling approaches. (shrink)

Intermediate constructs are required as bridges between complex behaviors and realistic models of neural circuitry. For cognitive scientists in general, schemas are the appropriate functional units; brain theorists can work with neural layers as units intermediate between structures subserving schemas and small neural circuits.After an account of different levels of analysis, we describe visuomotor coordination in terms of perceptual schemas and motor schemas. The interest of schemas to cognitive science in general is illustrated with the example of perceptual schemas (...) in high-level vision and motor schemas in the control of dextrous hands.Rana computatrix, the computational frog, is introduced to show how one constructs an evolving set of model families to mediate flexible cooperation between theory and experiment. Rana computatrix may be able to do for the study of the organizational principles of neural circuitry what Aplysia has done for the study of subcellular mechanisms of learning. Approach, avoidance, and detour behavior in frogs and toads are analyzed in terms of interacting schemas. Facilitation and prey recognition are implemented as tectal-pretectal interactions, with the tectum modeled by an array of tectal columns. We show how layered neural computation enters into models of stereopsis and how depth schemas may involve the interaction of accommodation and binocular cues in anurans. (shrink)

The proposed multilevel framework of discourse comprehension includes the surface code, the textbase, the situation model, the genre and rhetorical structure, and the pragmatic communication level. We describe these five levels when comprehension succeeds and also when there are communication misalignments and comprehension breakdowns. A computer tool has been developed, called Coh-Metrix, that scales discourse (oral or print) on dozens of measures associated with the first four discourse levels. The measurement of these levels with an automated tool helps researchers (...) track and better understand multilevel discourse comprehension. Two sets of analyses illustrate the utility of Coh-Metrix in discourse theory and educational practice. First, Coh-Metrix was used to measure the cohesion of the text base and situation model, as well as potential extraneous variables, in a sample of published studies that manipulated text cohesion. This analysis helped us better understand what was precisely manipulated in these studies and the implications for discourse comprehension mechanisms. Second, Coh-Metrix analyses are reported for samples of narrative and science texts in order to advance the argument that traditional text difficulty measures are limited because they fail to accommodate most of the levels of the multilevel discourse comprehension framework. (shrink)

An important lesson that philosophy can learn from the Turing Test and computer science more generally concerns the careful use of the method of Levels of Abstraction (LoA). In this paper, the method is first briefly summarised. The constituents of the method are “observables”, collected together and moderated by predicates restraining their “behaviour”. The resulting collection of sets of observables is called a “gradient of abstractions” and it formalises the minimum consistency conditions that the chosen abstractions must satisfy. Two useful (...) kinds of gradient of abstraction – disjoint and nested – are identified. It is then argued that in any discrete (as distinct from analogue) domain of discourse, a complex phenomenon may be explicated in terms of simple approximations organised together in a gradient of abstractions. Thus, the method replaces, for discrete disciplines, the differential and integral calculus, which form the basis for understanding the complex analogue phenomena of science and engineering. The result formalises an approach that is rather common in computer science but has hitherto found little application in philosophy. So the philosophical value of the method is demonstrated by showing how making the LoA of discourse explicit can be fruitful for phenomenological and conceptual analysis. To this end, the method is applied to the Turing Test, the concept of agenthood, the definition of emergence, the notion of artificial life, quantum observation and decidable observation. It is hoped that this treatment will promote the use of the method in certain areas of the humanities and especially in philosophy. (shrink)

Definition of the problem The umbrella term “explicability” refers to the reduction of opacity of artificial intelligence (AI) systems. These efforts are challenging for medical AI applications because higher accuracy often comes at the cost of increased opacity. This entails ethical tensions because physicians and patients desire to trace how results are produced without compromising the performance of AI systems. The centrality of explicability within the informed consent process for medical AI systems compels an ethical reflection on the trade-offs. Which (...) levels of explicability are needed to obtain informed consent when utilizing medical AI? Arguments We proceed in five steps: First, we map the terms commonly associated with explicability as described in the ethics and computer science literature, i.e., disclosure, intelligibility, interpretability, and explainability. Second, we conduct a conceptual analysis of the ethical requirements for explicability when it comes to informed consent. Third, we distinguish hurdles for explicability in terms of epistemic and explanatory opacity. Fourth, this then allows to conclude the level of explicability physicians must reach and what patients can expect. In a final step, we show how the identified levels of explicability can technically be met from the perspective of computer science. Throughout our work, we take diagnostic AI systems in radiology as an example. Conclusion We determined four levels of explicability that need to be distinguished for ethically defensible informed consent processes and showed how developers of medical AI can technically meet these requirements. (shrink)

Marr's seminal work laid out a program of research by specifying key questions for cognitive science at different levels of analysis. Because dynamic systems theory focuses on time and interdependence of components, DST research programs come to very different conclusions regarding the nature of cognitive change. We review a specific DST approach to cognitive-level processes: dynamic field theory. We review research applying DFT to several cognitive-level processes: object permanence, naming hierarchical categories, and inferring intent, that demonstrate the (...) difference in understanding of behavior and cognition that results from a DST perspective. These point to a central challenge for cognitive science research as defined by Marr—emergence. We argue that appreciating emergence raises questions about the utility of computational-level analyses and opens the door to insights concerning the origin of novel forms of behavior and thought. We contend this is one of the most fundamental questions about cognition and behavior. (shrink)

Recent attempts to develop and apply digital ethics principles to address the challenges of the digital transformation leave organisations with an operationalisation gap. To successfully implement such guidance, they must find ways to translate high-level ethics frameworks into practical methods and tools that match their specific workflows and needs. Here, we describe the development of a standardised risk assessment tool, the Principle-at-Risk Analysis (PaRA), as a means to close this operationalisation gap for a key level of the (...) ethics infrastructure at many organisations – the work of an interdisciplinary ethics panel. The PaRA tool serves to guide and harmonise the work of the Digital Ethics Advisory Panel at the multinational science and technology company Merck KGaA in alignment with the principles outlined in the company’s Code of Digital Ethics. We examine how such a tool can be used as part of a multifaceted approach to operationalise high-level principles at an organisational level and provide general requirements for its implementation. We showcase its application in an example case dealing with the comprehensibility of consent forms in a data-sharing context at Syntropy, a collaborative technology platform for clinical research. (shrink)

A study is described in which the effectiveness of a computer program (Hermes) on improving argumentative writing is tested. One group of students was randomly assigned to a control group and the other was assigned to the experimental group where they are asked to use the Hermes program. All students were asked to write essays on controversial topics to an opposed audience. Their essays were content-analysed for dialectical traits. Based on this analysis, it was concluded that the experimental group (...) wrote more dialectically effective essays than the control group, and the amount of difference between the control and experimental groups was related to the students' intellectual developmental level, as assessed by the Measure of Epistemological Reflection (MER). It is concluded that argumentative writing, operationalized here as dialectical writing, can be improved by computer-assisted instruction, but that attempts to teach such forms of thinking and writing need to take into account students' capacity to benefit from such instruction. Such capacity is defined here as intellectual development. (shrink)

We propose a critique of normativism, defined as the idea that human thinking reflects a normative system against which it should be measured and judged. We analyze the methodological problems associated with normativism, proposing that it invites the controversial “is-ought” inference, much contested in the philosophical literature. This problem is triggered when there are competing normative accounts (the arbitration problem), as empirical evidence can help arbitrate between descriptive theories, but not between normative systems. Drawing on linguistics as a model, we (...) propose that a clear distinction between normative systems and competence theories is essential, arguing that equating them invites an “is-ought” inference: to wit, supporting normative “ought” theories with empirical “is” evidence. We analyze in detail two research programmes with normativist features – Oaksford and Chater’s rational analysis and Stanovich and West’s individual differences approach – demonstrating how, in each case, equating norm and competence leads to an is-ought inference. Normativism triggers a host of research biases in the psychology of reasoning and decision making: focusing on untrained participants and novel problems, analyzing psychological processes in terms of their normative correlates, and neglecting philosophically significant paradigms when they do not supply clear standards for normative judgement. For example, in a dual-process framework, normativism can lead to a fallacious “ought-is” inference, in which normative responses are taken as diagnostic of analytic reasoning. We propose that little can be gained from normativism that cannot be achieved by descriptivist computational-level analysis, illustrating our position with Hypothetical Thinking Theory and the theory of the suppositional conditional. We conclude that descriptivism is a viable option, and that theories of higher mental processing would be better off freed from normative considerations. (shrink)

The general problem of visual search can be shown to be computationally intractable in a formal, complexity-theoretic sense, yet visual search is extensively involved in everyday perception, and biological systems manage to perform it remarkably well. Complexity level analysis may resolve this contradiction. Visual search can be reshaped into tractability through approximations and by optimizing the resources devoted to visual processing. Architectural constraints can be derived using the minimum cost principle to rule out a large class of potential (...) solutions. The evidence speaks strongly against bottom-up approaches to vision. In particular, the constraints suggest an attentional mechanism that exploits knowledge of the specific problem being solved. This analysis of visual search performance in terms of attentional influences on visual information processing and complexity satisfaction allows a large body of neurophysiological and psychological evidence to be tied together. (shrink)

There has been an ongoing conflict regarding whether reality is fundamentally digital or analogue. Recently, Floridi has argued that this dichotomy is misapplied. For any attempt to analyse noumenal reality independently of any level of abstraction at which the analysis is conducted is mistaken. In the pars destruens of this paper, we argue that Floridi does not establish that it is only levels of abstraction that are analogue or digital, rather than noumenal reality. In the pars construens of (...) this paper, we reject a classification of noumenal reality as a deterministic discrete computational system. We show, based on considerations from classical physics, why a deterministic computational view of the universe faces problems (e.g., a reversible computational universe cannot be strictly deterministic). (shrink)

Purpose The purpose of this study is to analyze how embedding of self-powered wireless sensors into cloud computing further enables such a system to become a sustainable part of work environment. Design/methodology/approach This is exemplified by an application scenario in healthcare that was developed in the context of the OpSIT project in Germany. A clearly outlined three-layer architecture, in the sense of Internet of Things, is presented. It provides the basis for integrating a broad range of sensors into smart healthcare (...) infrastructure. More specifically, by making use of short-range communication sensors, gateways which implement data transmission and low-level computation and cloud computing for processing the data. Findings A technical in-depth analysis of the first two layers of the infrastructure is given to prove reliability and to determine the communication quality and availability in real-world scenarios. Furthermore, two example use-cases that directly apply to a healthcare environment are examined, concluding with the feasibility of the presented approach. Practical implications Finally, the next research steps, oriented towards the semantic tagging and classification of data received from sensors, and the usage of advanced artificial intelligence-based algorithms on this information to produce useful knowledge, are described together with the derived social benefits. Originality/value The work presents an innovative, extensible and scalable system, proven to be useful in healthcare environments. (shrink)

In this paper we review some problems with traditional approaches for acquiring and representing knowledge in the context of developing user interfaces. Methodological implications for knowledge engineering and for human-computer interaction are studied. It turns out that in order to achieve the goal of developing human-oriented (in contrast to technology-oriented) human-computer interfaces developers have to develop sound knowledge of the structure and the representational dynamics of the cognitive system which is interacting with the computer.We show that in a first step (...) it is necessary to study and investigate the different levels and forms of representation that are involved in the interaction processes between computers and human cognitive systems. Only if designers have achieved some understanding about these representational mechanisms, user interfaces enabling individual experiences and skill development can be designed. In this paper we review mechanisms and processes for knowledge representation on a conceptual, epistemological, and methodologieal level, and sketch some ways out of the identified dilemmas for cognitive modeling in the domain of human-computer interaction. (shrink)

Based on a formal analysis of the operations Merge and Move, I provide a computational answer to the question why Move might be an integral part of language. The answer is rooted in the framework of subregular complexity, which reveals that Merge is most succinctly analyzed in terms of the formal class TSL. Any cognitive device that can handle this level of complexity also possesses sufficient resources for Move. In fact, Merge and Move are remarkably similar instances (...) of TSL. Consequently, Move has little computational or conceptual cost attached to it and comes essentially for free in any grammar that expresses Merge as compactly as possible. (shrink)

Increasingly, biologists are using computers to model and to create biological representations. However, the exponential growth in available biological dataposes a challenge for experimental and theoretical researchers in both Biology and in Computer Science. In short, when even the simple retrieval of relevant biological information for a researcher becomes a complex task — its analysis and synthesis with other biological information will become even more daunting and unlikely. In this context, specially organized ‘structures of representation’ are needed for the (...) efficient interpretation of experimentally generated data. The “semantic Web” is a recent trend in networking techniques that we will examine here as a possible strategy for the computation of biological data — one that may allow us to take into account both the semiotic dimensions of biological processes, as well as their dynamic organization into stable and systemic levels. Thereupon, we propose that a semantic network for biology could benefit from principles rooted in such previous representation efforts as computer-generated ‘landscapes’ and ‘physical attractors’ — and that such principles could, in turn, then be better integrated into biological research through the development of a more semiotically informed user / computer interface design. (shrink)

Intelligent algorithms together with various machine learning techniques hold a dominant position among major challenges for contemporary value sensitive design. Self-learning capabilities of current AI applications blur the causal link between programmer and computer behavior. This creates a vital challenge for the design, development and implementation of digital technologies nowadays. This paper seeks to provide an account of this challenge. The main question that shapes the current analysis is the following: What conceptual tools can be developed within the value (...) sensitive design school of thought for evaluating machine learning algorithms where the causal relation between designers and the behavior of their computer systems has been eroded? The answer to this question will be provided through two levels of investigation within the value sensitive design methodology. The first level is conceptual. Within the conceptual level, we will introduce the notion of computer intentionality and will show how this term may be used for solving an issue of non-causal relation between designer and computer system. The second level of investigation is technical. At this level the emphasis will be given to machine learning algorithms. (shrink)

The related fields of fractal image encoding and fractal image analysis have blossomed in recent years. This book, originating from a NATO Advanced Study Institute held in 1995, presents work by leading researchers. It is developing the subjects at an introductory level, but it also has some recent and exciting results in both fields. The book contains a thorough discussion of fractal image compression and decompression, including both continuous and discrete formulations, vector space and hierarchical methods, and algorithmic (...) optimizations. The book also discusses multifractal approaches to image analysis, segmentation, and recognition, including medical applications. (shrink)

This Ph.D. thesis explores a novel way of thinking about computation in cognitive science. It argues for what I call ‘the mechanistic account of wide computationalism’, or simply wide mechanistic computation. The key claim is that some cognitive and perceptual abilities are produced by or are the result of computational mechanisms that are, in part, located outside the individual; that computational systems, the ones that form the proper units of analysis in cognitive science, are particular types of (...) functional mechanisms that, on occasion, spread out across brain, body, and world. Wide mechanistic computation is the result of bringing together two distinct strands of thinking about computation: ‘wide’ views, which hold that computational systems can, on occasion, include parts of the surrounding environment; and ‘mechanistic’ views, which hold that computational explanation is a species of mechanistic explanation, and that computational mechanisms are a special type of functional mechanism. I argue that wide mechanistic computation draws support from several sources. First, I examine research on animal and human psychology and show that several organisms’ behaviours are properly treated as being the output of wide computational mechanisms. Second, I defend the view from several philosophical charges, including worries about its explanatory parsimony and empirical testability. Finally, I argue for the view’s theoretical credentials by showing that it can help resolve specific problems that have recently troubled 4E cognition. The result is an argument for not only the coherence but also empirical plausibility of wide mechanistic computation. On route to its main objective, the thesis also accomplishes a number of related tasks, including: providing a framework for organising and conceptualising different views of computation, securing the conceptual foundations of mechanistic computation by addressing an outstanding challenge called the ‘abstraction problem’, sounding a cautionary note about recent predictive processing accounts of extended cognition and arguing against a particular conception of levels often used within cognitive science, what is labelled the ‘hierarchical correspondence view of levels’. (shrink)

As China and the United States strive to be the primary global leader in AI, their visions are coming into conflict. This is frequently painted as a fundamental clash of civilisations, with evidence based primarily around each country’s current political system and present geopolitical tensions. However, such a narrow view claims to extrapolate into the future from an analysis of a momentary situation, ignoring a wealth of historical factors that influence each country’s prevailing philosophy of technology and thus their (...) overarching AI strategies. In this article, we build a philosophy-of-technology-grounded framework to analyse what differences in Chinese and American AI policies exist and, on a fundamental level, why they exist. We support this with Natural Language Processing methods to provide an evidentiary basis for our analysis of policy differences. By looking at documents from three different American presidential administrations––Barack Obama, Donald Trump, and Joe Biden––as well as both national and local policy documents (many available only in Chinese) from China, we provide a thorough comparative analysis of policy differences. This article fills a gap in US–China AI policy comparison and constructs a framework for understanding the origin and trajectory of policy differences. By investigating what factors are informing each country’s philosophy of technology and thus their overall approach to AI policy, we argue that while significant obstacles to cooperation remain, there is room for dialogue and mutual growth. (shrink)

According to Marr, explanations of perceptual behavior should address multiple levels of analysis. Rahnev & Denison (R&D) are perhaps overly dismissive of optimality considerations at the computational level. Also, an exclusive reliance on standard observer models may cause neglect of many other plausible hypotheses at the algorithmic level. Therefore, as far as explanation goes, standard observer modeling is no panacea.

The breadth-first search adopted by Bayesian researchers to map out the conceptual space and identify what the framework can do is beneficial for science and reflective of its collaborative and incremental nature. Theoretical pluralism among researchers facilitates refinement of models within various levels of analysis, which ultimately enables effective cross-talk between different levels of analysis.

Computational models of affect (CMAS), in their most common form, cannot take into account the qualitative (phenomenal) dimension of affect itself. Their expressivity can be extended, thus promoting the much sought-after standardization in the most theory-neutral way, using OWL (Web Ontology Language) and machine learning techniques. OWL is an expressive formal language, as well as an established open standard, and can be used to describe the models, possibly including qualitative entities at the fundamental level. The supervised machine learning (...) techniques allow the direct learning and application of models described as OWL ontologies. Thanks to human supervision (e.g. using datasets labeled by a human user), they can take into account the qualitative dimension of affect when the models warrant it. To further promote the aforementioned standardization, the task of classifying texts according to their affective content (known in computer science as "sentiment analysis") can be recommended as a way to assess the performance of the models. It is a multifaceted task, in which usually divergent fields as philosophy, psychology and computer science meet. Moreover, since it is a very current task in computer science, there are many resources available to facilitate the development of a standard benchmark for CMAS. (shrink)