Tzu-wei Hung

Artificial intelligence (AI) systems are increasingly adopted to make decisions in domains such as business, education, health care, and criminal justice. However, such algorithmic decision systems can have prevalent biases against marginalized social groups and undermine social justice. Explainable artificial intelligence (XAI) is a recent development aiming to make an AI system’s decision processes less opaque and to expose its problematic biases. This paper argues against technical XAI, according to which the detection and interpretation of algorithmic bias can be handled more or less independently by technical experts who specialize in XAI methods. Drawing on resources from feminist epistemology, we show why technical XAI is mistaken. Specifically, we demonstrate that the proper detection of algorithmic bias requires relevant interpretive resources, which can only be made available, in practice, by actively involving a diverse group of stakeholders. Finally, we suggest how feminist theories can help shape integrated XAI: an inclusive social-epistemic process that facilitates the amelioration of algorithmic bias.

This paper focuses on the potential of “equitech”—AI technology that improves equity. Recently, interventions have been developed to reduce the harm of implicit bias, the automatic form of stereotype or prejudice that contributes to injustice. However, these interventions—some of which are assisted by AI-related technology—have significant limitations, including unintended negative consequences and general inefficacy. To overcome these limitations, we propose a two-dimensional framework to assess current AI-assisted interventions and explore promising new ones. We begin by using the case of human resource recruitment as a focal point to show that existing approaches have exploited only a subset of the available solution space. We then demonstrate how our framework facilitates the discovery of new approaches. The first dimension of this framework helps us systematically consider the analytic information, intervention implementation, and modes of human-machine interaction made available by advancements in AI-related technology. The second dimension enables the identification and incorporation of insights from recent research on implicit bias intervention. We argue that a design strategy that combines complementary interventions can further enhance the effectiveness of interventions by targeting the various interacting cognitive systems that underlie implicit bias. We end with a discussion of how our cognitive interventions framework can have positive downstream effects for structural problems.

Should you be targeted by police for a crime that AI predicts you will commit? In this paper, we analyse when, and to what extent, the person-based predictive policing (PP) — using AI technology to identify and handle individuals who are likely to breach the law — could be justifiably employed. We first examine PP’s epistemological limits, and then argue that these defects by no means refrain from its usage; they are worse in humans. Next, based on major AI ethics guidelines (IEEE, EU, and RIKEN, etc.), we refine three basic moral principles specific to person-based PP. We also derive further requirements from case studies, including debates in Chicago, New Orleans, San Francisco, Tokyo, and cities in China. Instead of rejecting PP programs, we analyse what necessary conditions should be met for using the tool to achieve social good. While acknowledging its risks, we conclude that the person-based PP could be beneficial in community policing, especially when merging into a larger governance framework of the social safety net.

This paper investigates the representation of authors with different linguistic backgrounds in academic publishing. We first review some common rebuttals of concerns about linguistic injustice. We then analyze 1039 authors of philosophy journals, primarily selected from the 2015 Leiter Report. While our data show that Anglophones dominate the output of philosophy papers, this unequal distribution cannot be solely attributed to language capacities. We also discover that ethics journals have more Anglophone authors than logic journals and that most authors are affiliated with English-speaking universities, suggesting other factors may also play significant roles. Moreover, some interesting results are revealed when we combine the factor of sex with place of affiliation and linguistic background. It indicates that while certain linguistic injustice is inevitable in academic publishing, it may be more complex than thought. We next introduce Broadbent’s :457–476, 2012, Philosophy of epidemiology, Palgrave Macmillan, New York, 2013, Stud Hist Philos Biol Biomed Sci 48:250–257, 2014) contrastive account of causation to give a causal explanation of our findings. Broadbent’s account not only well characterizes the multifaceted causality in academic publishing but also provides a methodological guideline for further investigation.

This paper examines racial discrimination and algorithmic bias in predictive policing algorithms (PPAs), an emerging technology designed to predict threats and suggest solutions in law enforcement. We first describe what discrimination is in a case study of Chicago’s PPA. We then explain their causes with Broadbent’s contrastive model of causation and causal diagrams. Based on the cognitive science literature, we also explain why fairness is not an objective truth discoverable in laboratories but has context-sensitive social meanings that need to be negotiated through democratic processes. With the above analysis, we next predict why some recommendations given in the bias reduction literature are not as effective as expected. Unlike the cliché highlighting equal participation for all stakeholders in predictive policing, we emphasize power structures to avoid hermeneutical lacunae. Finally, we aim to control PPA discrimination by proposing a governance solution—a framework of a social safety net.

The language parasite approach refers to the view that language, like a parasite, is an adaptive system that evolves to fit its human hosts. Supported by recent computer simulations, LPA proponents claim that the reason that humans can use languages with ease is not because we have evolved with genetically specified linguistic instincts but because languages have adapted to the preexisting brain structures of humans. This article examines the LPA. It argues that, while the LPA has advantages over its rival, Chomskyan nativism, there are additional factors that may limit linguistic variety that have yet to be identified by its insightful proponents. This article suggests abandoning the search for a decisive cause of language capacity and argues that language evolution is more likely to arise from balancing multiple engineering constraints.

Whereas using artificial intelligence (AI) to predict natural hazards is promising, applying a predictive policing algorithm (PPA) to predict human threats to others continues to be debated. Whereas PPAs were reported to be initially successful in Germany and Japan, the killing of Black Americans by police in the US has sparked a call to dismantle AI in law enforcement. However, although PPAs may statistically associate suspects with economically disadvantaged classes and ethnic minorities, the targeted groups they aim to protect are often vulnerable populations as well (e.g., victims of human trafficking, kidnapping, domestic violence, or drug abuse). Thus, determining how to enhance the benefits of PPA while reducing bias through better management is important. In this paper, we propose a policy schema to address this issue. First, after clarifying relevant concepts, we examine major criticisms of PPAs and argue that some of them should be addressed. If banning AI or making it taboo is an unrealistic solution, we must learn from our errors to improve AI. We next identify additional challenges of PPAs and offer recommendations from a policy viewpoint. We conclude that the employment of PPAs should be merged into broader governance of the social safety net and audited publicly by parliament and civic society so that the unjust social structure that breeds bias can be revised.

Cognitive warfare—controlling others’ mental states and behaviors by manipulating environmental stimuli—is a significant and ever-evolving issue in global conflict and security, especially during the COVID-19 crisis. In this article, we aim to contribute to the field by proposing a two-dimensional framework to evaluate China's cognitive warfare and explore promising ways of counteracting it. We first define the problem by clarifying relevant concepts and then present a case study of China's attack on Taiwan. Next, based on predictive coding theory from the cognitive sciences, we offer a framework to explain how China's cognitive warfare works and to what extent it succeeds. We argue that this framework helps identify vulnerable targets and better explains some of the conflicting data in the literature. Finally, based on the framework, we predict China's strategy and discuss Taiwan's options in terms of cognitive and structural interventions.

Despite intensive debates regarding action imitation and sentence imitation, few studies have examined their relationship. In this paper, we argue that the mechanism of action imitation is necessary and in some cases sufficient to describe sentence imitation. We first develop a framework for action imitation in which key ideas of Hurley’s shared circuits model are integrated with Wolpert et al.’s motor selection mechanism and its extensions. We then explain how this action-based framework clarifies sentence imitation without a language-specific faculty. Finally, we discuss the empirical support for and philosophical significance of this perspective

Although the relationship between prejudice and predictive coding has attracted more attention recently, many important issues remain to be investigated, such as why prejudice is so persistent and how to accommodate seemingly conflicting studies. In this paper, we offer an integrated framework to explain the functional-computational mechanism of prejudice. We argue that this framework better explains (i) why prejudice is somewhat immune to revision, (ii) how inconsistent processing (e.g. one’s moral belief and biased emotional reaction) may occur, (iii) the dispute over the content of ‘alief’ (unconscious belief), and (iv) conflicting data reported in disinformation studies. Finally, we briefly review the main intervention strategies and offer recommendations.

This article systematically surveys the history of Anglo-American philosophy in Taiwan since the late nineteenth century. Contrary to conventional wisdom, it argues that pragmatism remained influential given the dominance of continental philosophy in Japanese colonized Taiwan, where the universal values assumed by pragmatists were used to resist the Empire’s ideology, after WWII, immigrated Chinese scholars brought in more novelty to Taiwanese philosophy than the Vienna circle diasporas brought to their Anglo-American counterparts, in which liberal scholars’ emphasis on science and democracy challenged Chiang’s dictatorship and Chinese nationalism, and analytic philosophy thrived with the advance of democratizing in the 1980s, which, in turn, diversified the philosophy in 1990s Taiwan. Accordingly, Anglo-American philosophy was never confined to the academic ivory tower but had positive impacts on Taiwan’s centennial social transformation.

Philosophy in Taiwan can be traced to the age of discovery, when Dutch colonists preached the monotheism of Calvinism.1 This was followed by the introduction of Neo-Confucianism by Koxinga's regime. Buddhism and Daoism, and, later, Scottish Presbyterianism, were imported during the Manchu Qing period. Starting in 1895, Japan significantly modernized its first colony in Taiwan, bringing in European philosophy and the Kyoto School. When the communists took over China in 1949, the best liberal and Confucian scholars fled to Taiwan, which has imported analytic philosophy since the 1960s. Recently, epistemology and environmental ethics based on Austronesian...

"Rationality: Contexts and Constraints" is an interdisciplinary reappraisal of the nature of rationality. In method, it is pluralistic, drawing upon the analytic approaches of philosophy, linguistics, neuroscience, and more. These methods guide exploration of the intersection between traditional scholarship and cutting-edge philosophical or scientific research. In this way, the book contributes to development of a suitably revised, comprehensive understanding of rationality, one that befits the 21st century, one that is adequately informed by recent investigations of science, pathology, non-human thought, emotion, and even enigmatic Chinese texts that might previously have seemed to be expressions of irrationalism. Addresses recent challenges and Identifies a direction for future research on rationalityInvestigates the relationship between rationality and mental disorders, such as delusion and depressionAssesses reasoning in artificial intelligence and nonhuman animalsReflects on ancient Chinese Philosophy and possible cultural differences in human psychology Employs philosophical reflection, along with linguistic, probabilistic, and logical techniques.

Hung Yao-hsün is one of the most creative, albeit long overlooked, thinkers in Japanese-ruled Taiwan. This paper’s aim is threefold. It first argues that while Hung’s early philosophy was rooted in the Kyoto school, he is a key founder of the Sit-chûn movement of Taiwanese philosophy. It next shows that during Taiwan’s martial law, Hung’s thought features a “Buddhist turn,” in which Zen is incorporated within existentialism. Third, while this turn is a sharp contrast to his prewar philosophical activism, Hung’s last work stressed Abraham Kaplan’s view that philosophy should be connected to one’s life experience, echoing Hung’s prewar usage of fūdo in justifying Taiwan’s cultural subjectivity. In other words, there is an implicit continuity between his early and late philosophy.

This book focuses on the connection between action and verbal communication, exploring topics such as the mechanisms of language processing, action processing, voluntary and involuntary actions, knowledge of language and assertion. Communication modelling and aspects of communicative actions are considered, along with cognitive requirements for nonverbal and verbal communicative action. Contributions from expert authors are organised into three parts in this book, focussing on language in communication, action and bodily awareness, and sensorimotor interaction and language acquisition. Readers will discover various methods that have been employed in investigations presented here, including neurological experiment, computational modeling, and logical and philosophical analysis. These diverse expert perspectives shed light on the extent to which a mechanism for processing actions also facilitates the processing of language, and the authors' work prompts further interdisciplinary investigation of the relationship between language and action. This book is written for readers from different academic backgrounds; from graduate students to established academics in disciplines ranging from neuroscience to psychology, philosophy, linguistics and beyond. Earlier versions of the selected essays in this book were presented at the 2013 IEAS Conference on Language and Action, held in Taipei, Taiwan.

The enzyme model, inspired by biological enzyme catalysis, is a computational-functional description of information processing and distribution in modular cognitive systems. It has been argued that EM offers advantages in solving both the allocation problem and global computation and thus may play a role in upholding the massive modularity hypothesis. This paper, however, argues that EM solutions are untenable, as EM avoids the infinite regress of allocation problem only at a high cost and with several critical drawbacks. Moreover, to clarify global processes, EM needs to satisfy two necessary conditions: first to demonstrate that the EM allows cross module communication, and second to be sensitive to not only the syntax but also the semantics of representations. I argue that EM only satisfies the first condition and thus fails to hold.