Generalizable Explanations Research Articles

Explaining any black box model using real data

In recent years the use of complex machine learning has increased drastically. These complex black box models trade interpretability for accuracy. The lack of interpretability is troubling for, e.g., socially sensitive, safety-critical, or knowledge extraction applications. In this paper, we propose a new explanation method, SLISE, for interpreting predictions from black box models. SLISE can be used with any black box model (model-agnostic), does not require any modifications to the black box model (post-hoc), and explains individual predictions (local). We evaluate our method using real-world datasets and compare it against other model-agnostic, local explanation methods. Our approach solves shortcomings in other related explanation methods by only using existing data instead of sampling new, artificial data. The method also generates more generalizable explanations and is usable without modification across various data domains.

Towards a general model explaining physical and digital counterfeits

PurposeDemand for fake physical and digital products is a global phenomenon with substantive detrimental effects on companies and consumers. This raises various questions and issues, such as whether there are generalizable explanations of purchase intentions.Design/methodology/approachThis research is based on consumer samples from three different countries. This paper develops and tests a model based on the theory of planned behavior (TPB) to explain both the demand for counterfeits and digital piracy. Respondents were questioned about physical products (e.g. clothing, accessories) from well-known brands and digital products (e.g. software, music).FindingsSocially oriented motives such as embarrassment potential, ethical concerns and social norms explain the intention to purchase fake physical and digital products, while personally oriented motives (e.g. self-identity) have indirect effects but not a direct impact on purchase intention.Research limitations/implicationsAs our results show, we find evidence for a general model – contributing and supporting our first and primary research goal of providing a theoretically robust model that bridges the gap between two streams of literature.Practical implicationsThe fact that drivers of buying counterfeit physical and digital goods are similar across countries provides justification for companies and international organizations to bundle their efforts and thus leverage them more strongly on a global scale.Originality/valueWe provide a basis for consolidating future research on demand for counterfeits and pirated goods because underlying factors driving demand are similar across the three countries studied herein.

When is templatic morphology borrowed?

Semitic languages are typologically unusual in making extensive morphological use of so-called “root-and-pattern” morphology, in the form of fixed-length templates that fix vowel qualities in the output while ignoring the vowels of the input. The expansion of Arabic over the past 1500 years has created ideal conditions for the borrowing of fixed-vowel fixed-length templates into the languages of massively bilingual minority groups in the Arab world. Prominent among the morphemes borrowed in such circumstances is the comparative/superlative template ʔaCCaC, conventionally termed the elative. This template has become fully productive in languages including Siwi Berber, Western Neo-Aramaic, and Mehri, and suppletively productive in Domari. A nearly exhaustive examination of massively bilingual minority groups in the Arab world suggests that the outcome is determined not only by sociolinguistic factors but also by structural ones: only languages with pre-existing triliteral fixed-vowel templates – used in particular for change-of-state verbs – borrow this template in a fully productive fashion, while other languages, if they borrow it at all, are forced to resort to suppletion and/or to leave it unproductive. This observation is consistent with two more broadly generalisable explanations: that the productive borrowing of “root-and-pattern” morphology requires not only the borrowing of its outputs but also the presence (through borrowing or common inheritance) of enough of the corresponding inputs, and that, in any given category, pre-existing root extraction processes are a precondition for the productive borrowing of “root-and-pattern” morphology.

The soil priming effect: Consistent across ecosystems, elusive mechanisms

Organic matter input to soils can accelerate the decomposition of native soil carbon (C), a process called the priming effect. Priming is ubiquitous and exhibits some consistent patterns, but a general explanation remains elusive, in part because of variation in the response across different ecosystems, and because of a diversity of proposed mechanisms, including microbial activation, stoichiometry, and community shifts. Here, we conducted five-week incubations of four soils (grassland, piñon-juniper, ponderosa pine, mixed conifer), varying the amount of substrate added (as 13C-glucose, either 350 or 1000 μg C g−1 week−1) and either with no added nitrogen (N), or with sufficient N (as NH4NO3) to bring the C-to-N ratio of the added substrate to 10. Using four different ecosystems enabled testing the generality of mechanisms underlying the priming effect. The responses of priming to the amount and C-to-N ratio of the added substrate were consistent across ecosystems: priming increased with the rate of substrate addition and declined when the C-to-N ratio of the substrate was reduced. However, structural equation models failed to confirm intermediate responses postulated to mediate the priming effect, including responses postulated to be mediated by stoichiometry and microbial activation. Specifically, priming was not clearly associated with changes in microbial biomass or turnover, nor with extracellular enzyme activities or the microbial C-to-N ratio. The strongest explanatory pathways in the structural equation models were the substrate, soil, and C-to-N ratio treatments themselves, with no intermediates, suggesting that either these measurements lacked sufficient sensitivity to reveal causal relationships, or the actual drivers for priming were not included in the ancillary measurements. While we observed consistent changes in priming caused by the amount and C-to-N ratio of the added substrate across a wide array of soils, our findings did not clearly conform to common models offered for the priming effect. Because priming is a residual flux involving diverse substrates of varying chemical composition, a simple and generalizable explanation of the phenomenon may be elusive.

Needs-based provision of medical care to nursing home residents: protocol for a mixed-methods study

IntroductionNursing home residents typically have greater needs for medical care than community-dwelling elderly. However, restricted cognitive abilities and limited mobility may impede their access to general practitioners and medical specialists....

Identification of Causal Effects in the Presence of Selection Bias

Cause-and-effect relations are one of the most valuable types of knowledge sought after throughout the data-driven sciences since they translate into stable and generalizable explanations as well as efficient and robust decision-making capabilities. Inferring these relations from data, however, is a challenging task. Two of the most common barriers to this goal are known as confounding and selection biases. The former stems from the systematic bias introduced during the treatment assignment, while the latter comes from the systematic bias during the collection of units into the sample. In this paper, we consider the problem of identifiability of causal effects when both confounding and selection biases are simultaneously present. We first investigate the problem of identifiability when all the available data is biased. We prove that the algorithm proposed by [Bareinboim and Tian, 2015] is, in fact, complete, namely, whenever the algorithm returns a failure condition, no identifiability claim about the causal relation can be made by any other method. We then generalize this setting to when, in addition to the biased data, another piece of external data is available, without bias. It may be the case that a subset of the covariates could be measured without bias (e.g., from census). We examine the problem of identifiability when a combination of biased and unbiased data is available. We propose a new algorithm that subsumes the current state-of-the-art method based on the back-door criterion.

The politics of uneven smallholder cacao expansion: A critical physical geography of agricultural transformation in Southeast Sulawesi, Indonesia

This article reconstructs and explains variability in pathways of smallholder cacao expansion and land use and cover change (LUCC) over the past four decades in Southeast Sulawesi, Indonesia. I first draw on approaches from remote sensing (RS) to reconstruct cacao expansion in two top-producing districts (1972–2014), highlighting the variegated environment changes shaped by smallholder cacao plantings, both inside and outside forested lands. I then integrate these data with theories and methods from political ecology (PE) and critical physical geography (CPG) to document the uneven politics of land and capital access mediating this variability. Using these data, I argue that variability in crop expansion and LUCC should not be considered as an exception but as the norm; integral to any generalizable explanation of LUCC. I also show how a focus on variability can not only supplement but also shift dominant accounts of LUCC.

Interpretation and Social Explanation

In this article I focus on those aspects of Keith Dowding’s book that are most concerned with interpretive approaches to the study of politics. I argue that, in ways not adequately captured by Dowding’s descriptions, the historical study of political concepts tells us something about their historical political effects and for this reason has a distinct value for how we think about and study politics. Furthermore, I argue, concepts of and about politics, including the concepts of political science, cannot be fully separated from the political contexts of which they are a part. Concepts which function as generalisable explanations at one point in time can shape the thinking and behaviour of political actors and thus be very particular causes. A philosophy or method of political science unaware of or inattentive to this dimension of politics and political science is incomplete.

Research highlights for issue 10: understanding complex lifecycles.

Many parasites, including those of relevance to human health, use multiple hosts in order to complete their lifecycle. These complex lifecycles are somewhat mysterious from an evolutionary perspective, as the reliance on more than one host species seems likely to make the parasite more vulnerable to ecosystem perturbation and to restrict its range (Lopez et al. 2015). There have been a number of intriguing hypotheses put forward to explain how selection could favor such strategies – ranging from somewhat neutral explanations, such as selection to simply survive predation of hosts, to elaborately adaptive ones, such as selection to exploit hosts of larger size by moving up the food chain (Parker et al. 2015a; Poulin and Lagrue 2015). However, finding evidence to support these hypotheses – and especially to craft generalizable explanations – has proved difficult.

Past Minds: Studies in Cognitive Historiography, written by Luther H. Martin and Jesper Sørensen

Luther H. Martin and Jesper Sorensen (eds.) Past Minds: Studies in Cognitive Historiography (Religion, Cognition, and Culture). London: Equinox, 2011. xiii + 206 pp. ISBN 9781845537418 (pbk.) This is the second volume in the Religion, Cognition, and Culture series edited by Jeppe Sinding Jensen and Armin W. Geertz. As noted in the preface, this volume, edited by Luther H. Martin and Jesper Sorensen, emerged from an international symposium on the theme of “Past Minds: Evolution, Cognition, and History,” which was held at the Institute of Cognition and Culture, Queen’s University Belfast in May 2007. The volume is comprised of four sections with four theoretical papers on cognitive historiography in the introductory and concluding sections and nine case studies, drawn from ancient religions, in two sections devoted respectively to the Roman world and various other ancient civilizations. The papers are linked by an overarching interest in the transmission of tradition and the cognitive mechanisms that underwrite that process. Overall, the volume makes a case for the value of cognitive and evolutionary theories for explaining history at different levels of historical analysis. In the first of the four theoretical chapters, Luther Martin provides a historical overview of the relationship between historiography and evolutionary theorizing in the wake of Darwin’s On the Origin of the Species, and of the search for generalizable explanations of historical events. In the second, Christophe Heintz discusses the importance of cultural epidemiology as a theoretical framework that allows historians to analyze and explain the distribution of cultural representations and material cultural forms. In the penultimate chapter, Don Wiebe worries that the Integrated Causal Model advanced in The Adapted Mind: Evolutionary Psychology and the Generation of Culture (1992), the edited volume by Jerome H. Barkow, Leda Cosmides, and John Tooby, may not leave room for complementary approaches and makes a case for explanatory pluralism. Sorensen’s concluding essay provides an analysis, first, of three potential ways of conceiving the relationship between history (the attempt to understand the relations between particular events), socio- cultural systems (the modeling of stable modes of social organization and public symbolic representations), and psychology (understood as the modeling of individuals’ neurocognitive systems) and, second, of the role of cognitive theorizing at three different levels of historical inquiry, which he refers to as micro, macro, and meso. Tooby and Cosmides’s widely-read chapter on “The Psychological Foundations of Culture” (1992); Dan Sperber’s Explaining Culture (1996); and to a lesser but still significant extent, Robert Boyd and Peter J. Richerson’s Culture and the Evolutionary Process (1985) figure prominently in all four theoretical chapters, which provide a convenient point of entry into the case studies included in this volume. Our review, therefore, primarily focuses on theoretical issues pertaining to the underlying discussions of evolutionary psychology, cultural epidemiology, and gene-culture co-evolution in the theoretical chapters and more briefly with the individual case studies.

Evaluating Three Explanations for the Design of Bilateral Investment Treaties

Although many features of bilateral investment treaties (BITs) are consistent from one agreement to the next, a closer look reveals that the treaties exhibit considerable variation in terms of their enforcement provisions, which legal scholars have singled out as the central component of the treaties. An original data set is compiled that captures three important treaty-design differences: whether the parties consent in advance to international arbitration, whether they allow treaty obligations to be enforced before an institutionalized arbitration body, and how many arbitration options are specified for enforcement. Drawing upon several relevant literatures on international institutions, three potentially generalizable explanations for this important treaty variation are articulated and tested. The strongest support is found for the theoretical perspective that emphasizes the bargaining power and preferences of capital-exporting states, which use the treaties to codify strong, credible investor protections in all their treaties. Empirical tests consistently reveal that treaties contain strong enforcement provisions—in which the parties preconsent to multiple, often institutionalized arbitration options—when the capital-exporting treaty partner has considerable bargaining power and contains domestic actors that prefer such arrangements, such as large multinational corporations or right-wing governments. In contrast, there is no evidence to support the popular hands-tying explanation, which predicts that investment-seeking states with the most severe credibility problems, due to poor reputations or weak domestic institutions, will bind themselves to treaties with stronger investment protections. likewise, little support is found for explanations derived from the project on the rational design of international institutions, which discounts the identities and preferences of the treaty partners and instead emphasizes the structural conditions they jointly face. In sum, this foundational study of differences across investment treaties suggests that the design of treaties is driven by powerful states, which include elements in the treaties that serve their interests, regardless of the treaty partner or the current strategic setting.

Exploring action potential initiation in neurons exposed to DC electric fields through dynamical analysis of conductance-based model

Noninvasive direct current (DC) electric stimulation of central nervous system is today a promising therapeutic option to alleviate the symptoms of a number of neurological disorders. Despite widespread use of this noninvasive brain modulation technique, a generalizable explanation of its biophysical basis has not been described which seriously restricts its application and development. This paper investigated the dynamical behaviors of Hodgkin’s three classes of neurons exposed to DC electric field based on a conductance-based neuron model. With phase plane and bifurcation analysis, the different responses of each class of neuron to the same stimulation are shown to derive from distinct spike initiating dynamics. Under the effects of negative DC electric field, class 1 neuron generates repetitive spike through a saddle-node on invariant circle (SNIC) bifurcation, while it ceases this repetitive behavior through a Hopf bifurcation; Class 2 neuron generates repetitive spike through a Hopf bifurcation, meanwhile it ceases this repetitive behavior also by a Hopf bifurcation; Class 3 neuron can generate single spike through a quasi-separatrix-crossing (QSC) at first, then it generates repetitive spike through a Hopf bifurcation, while it ceases this repetitive behavior through a SNIC bifurcation. Furthermore, three classes of neurons’ spiking frequency f–electric field E (f–E) curves all have parabolic shape. Our results highlight the effects of external DC electric field on neuronal activity from the biophysical modeling point of view. It can contribute to the application and development of noninvasive DC brain modulation technique.

The INGO Research Agenda: A Community Approach to Challenges in Method and Theory

IR scholars have demonstrated the significance of INGOs to international relations, but we have yet to establish more generalizable explanations and theories of why, and under what conditions, INGOs affect the interests and behavior of other actors or create institutional change.

Realism at the Limits: Post-Cold War Realism and Nuclear Rollback

Realists have been all but silent on the issue of nuclear disarmament, largely because giving up military capabilities seems counter-intuitive to realism's logic. The nonproliferation literature, on the other hand, has treated the insights of realist theory as failing to come to grips with the complex nature of a state's decision-making processes. The non-proliferation literature has produced rich empirical details of nuclear rollback; however, much of this literature lacks a general theoretical framework to provide generalizable explanations and predictions. This article advances two defensive realist hypotheses on state behaviour to explain nuclear rollback. It argues that states may voluntarily reduce their own capabilities in order to survive in an anarchic order when, in the absence of a secure second-strike capability, they make the rational decision to give up their nuclear arsenals. However, the absence of a secure second-strike capability is a necessary but not sufficient condition for nuclear disarmament. The absence of a dyadic existential threat is also a necessary condition for a state to give up its nuclear weapons. Together these two conditions, from a defensive realist perspective, are jointly necessary and sufficient for nuclear disarmament. The findings suggest that incentives for disarmament, from a realist perspective, can be made by taking into account the strategic interaction that states find themselves in vis-à-vis their environment and capabilities.

Biophysical Basis for Three Distinct Dynamical Mechanisms of Action Potential Initiation

Transduction of graded synaptic input into trains of all-or-none action potentials (spikes) is a crucial step in neural coding. Hodgkin identified three classes of neurons with qualitatively different analog-to-digital transduction properties. Despite widespread use of this classification scheme, a generalizable explanation of its biophysical basis has not been described. We recorded from spinal sensory neurons representing each class and reproduced their transduction properties in a minimal model. With phase plane and bifurcation analysis, each class of excitability was shown to derive from distinct spike initiating dynamics. Excitability could be converted between all three classes by varying single parameters; moreover, several parameters, when varied one at a time, had functionally equivalent effects on excitability. From this, we conclude that the spike-initiating dynamics associated with each of Hodgkin's classes represent different outcomes in a nonlinear competition between oppositely directed, kinetically mismatched currents. Class 1 excitability occurs through a saddle node on invariant circle bifurcation when net current at perithreshold potentials is inward (depolarizing) at steady state. Class 2 excitability occurs through a Hopf bifurcation when, despite net current being outward (hyperpolarizing) at steady state, spike initiation occurs because inward current activates faster than outward current. Class 3 excitability occurs through a quasi-separatrix crossing when fast-activating inward current overpowers slow-activating outward current during a stimulus transient, although slow-activating outward current dominates during constant stimulation. Experiments confirmed that different classes of spinal lamina I neurons express the subthreshold currents predicted by our simulations and, further, that those currents are necessary for the excitability in each cell class. Thus, our results demonstrate that all three classes of excitability arise from a continuum in the direction and magnitude of subthreshold currents. Through detailed analysis of the spike-initiating process, we have explained a fundamental link between biophysical properties and qualitative differences in how neurons encode sensory input.

Consumer Brand Equity in a Cross-cultural Replication: An Evaluation of a Scale

This paper aims to contribute positively to the now growing 'replicated studies' research. The replication of published empirical findings is an important means of developing robust and generalisable explanations in brand management. The purpose of this is to replicate the consumer-based equity scale developed by Vazquez, Rio and Iglesias (2002) and to determine whether the scale could be applied to a different cultural setting. To achieve this, our research will: (1) replicate the four main dimension of Vazquez et. al. scale (i.e. product, functional utility, product symbolic utility, brand name functional utility, brand name symbolic utility); (2)identify the similarities and differences of the application of this scale in Turkey (a culture which is significantly different from that for which the scale was initially developed). Confirmatory factor analysis was used and the key findings from the study indicated that, although the scale fits the Turkish culture, it needed small changes.

The Perils of Profiling: Civil War Spoilers and the Collapse of Intrastate Peace Accords

In a highly influential article in International Security, Stephen Stedman introduced a model of “civil war spoilers,” which focused valuable attention on the generally underappreciated role of elites in determining the course of negotiations and in implementing intrastate peace accords. For all its virtues, however, the spoiler model did not suggest the best set of strategies for deterring or defeating those who might seek to undermine peace processes. This is because context-specific and actor-specific measures tend to affect diplomatic instruments only at the margin and because, while spoiler type does not change over time, actors' commitment to fulfilling the provisions of peace accords often does; thus these static characteristics cannot be the critical variables the spoiler model suggests they are. Instead, as a detailed reexamination of three of Stedman's case studies (i.e., Angola, Mozambique, and Cambodia) demonstrates, a capabilities-based model offers a more parsimonious and generalizable explanation for why, when, and under what conditions actors who seek to undermine the peace will emerge or retreat. As such, the real key to deterring and defeating would-be spoilers lies in the possession and exercise of the material power to coerce or co-opt them, rather than in the capacity to discern their true character or personality type.

A Multiple Streams Model of U.S. Foreign Aid Policy

This study builds an interactive, quantitative model to explain U.S. foreign aid allocations. We develop an explanation based on Kingdon's multiple streams model to argue that aid decisions are the result of trade ties, socialist orientation, human needs, and the political ideologies of the administration and the senate interacting with adjustments to baseline funding. We find that there are significant conditional relationships between economic and security‐related assistance and our external and domestic variables with external variables generally being more important. The findings have implications for the study of public policy, the future of the foreign aid program, and theoretical attempts to develop more generalizable explanations of policy that encompass foreign and domestic issues.

Solving Methodological Challenges Using a Theory-driven Evaluation in the Study of Complex Patient Care

This article draws on two research designs, which were used in a study of the palliative management of malignant wounds, to demonstrate the inherent challenges in producing generalizable knowledge from a complex subject. The designs included quasi-experimentation and a form of theory- driven evaluation. The focus of the study was the performance of dressings to reduce the impact of the wounds on daily life. The convention for evaluating the outcomes of interventions is a controlled study design. This approach is considered the valid way of producing evidence that is generalizable. An experimental design was therefore adopted to conduct a series of n = 1 quasi-experiments on wound dressing performance with qualitative methods to explore individual experiences. Problems with the methods, in particular the failure to capture complex inter-relationships between clinical problems and patient experiences, forced the researcher to abandon the experimental design. An alternative approach to measurement was adopted to maintain the study focus, the TELER® system of treatment evaluation, which includes indicators of patient-centred outcomes of care. The philosophical position of the study was reviewed. Consensus emerged as an unalterable, major in uence on the design and paradigmatic assumptions of the study. A `system of reasoning' was adopted to overcome the inherent relativist position of knowledge derived from this approach. The system abstracts general issues from case study data to construct theoretical explanations that may be consistent with, or challenge, current knowledge. This article is based on the evolution of one particular study. However, it makes a more general contribution to evaluation research by explaining the rationale for a form of theory-driven evaluation that uses evidence, reason and theory to develop generalizable explanations from complex, individual case study data. The methodology is proposed for other complex situations where speci c and rigorous evidence, capable of generalization is needed.

Working with difference differently: exploring the subtext of pedagogical interactions

As a white feminist university teacher, I am interested in thinking about how to work with difference differently. In this article I will argue that part of what is required is the development of communicative capacities and relationships which attend to the intersubjective and unconscious aspects of pedagogical interactions. Focusing on the ways emotionally charged moments in classroom interactions are formed by, embedded in and reflective of particular social, emotional, and discursive histories, my article attempts to delineate a more complex understanding of what is at stake in classroom interactions in order to open spaces to create possibilities for ‘different grammars’ within academic sites. My intention, in focusing on moments when the apparent order of the classroom and its participants is disrupted or unsettled in some way, is not to develop generalisable explanations or solutions, let alone ways to ‘overcome’ the challenges such moments present to teachers and students. Although moments of crisis in classrooms may be familiar, the specificities of the conjunctures informing such moments are always situated and highly specific. My argument is that psychoanalytic theory offers resources for working with and learning from the specificities of these moments.