New Foundations for the Study of Biological and Artificial Intelligence

FAIN: FEL-282501-22

Lisa Titus
Colorado Seminary (Denver, CO 80210-4711)

Research and writing of a book on the distinction between artificial and human intelligence.

The primary deliverable will be an academic monograph, "Explaining Intelligence: New Foundations for the Study of Biological and Artificial Intelligence," which will provide much-needed philosophical foundations for understanding the nature of intelligence, elucidating and advancing its empirical study, and understanding its social and ethical impacts. The PI will develop and support a novel, non-reductive approach to the study of intelligence centered on minded intelligence: the kind of intelligence that beings with minds (including humans and many non-human animals, but not current AI systems) have. This work will demonstrate the central importance of humanistic inquiries into the complexities of studying intelligence in cognitive science, AI, and, and robotics contexts.

Associated Products

Embodied Cognition and the Causal Roles of the Mental (Book Section)
Title: Embodied Cognition and the Causal Roles of the Mental
Author: Lisa Miracchi Titus
Editor: Michael Brent
Editor: Lisa Miracchi Titus
Abstract: Embodied cognition theorists who argue that cognition is grounded in, or constituted by, bodily and environmental processes face the challenge of explaining why the body and environment are not merely causally, but metaphysically, relevant (Block 2005, Aizawa 2007). This is especially pressing because much of the support for this embodied cognition claim comes from empirical evidence for the causal interrelatedness of bodily, environmental, and neural processes, which at least prima facie could equally support both hypotheses. I provide a novel response to this challenge. I argue that if we are accept the widely adopted thesis of Semantic Externalism -- namely that the contents of our mental states are metaphysically determined by features beyond our brains -- and if we also accept the plausible thesis of Semantic Efficacy -- the thesis that the contents of our mental states are causally relevant to how we think and act -- then we must accept that features beyond our brains make a metaphysical difference to mental processes.
Year: 2023
Primary URL: https://www.routledge.com/Mental-Action-and-the-Conscious-Mind/Brent-Titus/p/book/9780367077518
Access Model: Subscription only
Publisher: Routledge
Book Title: Mental Action and the Conscious Mind

Mental Action and the Conscious Mind (Book)
Title: Mental Action and the Conscious Mind
Editor: Michael Brent
Editor: Lisa Miracchi Titus
Abstract: Mental action deserves a place among foundational topics in action theory and philosophy of mind. Recent accounts of human agency tend to overlook the role of conscious mental action in our daily lives, while contemporary accounts of the conscious mind often ignore the role of mental action and agency in shaping consciousness. This collection aims to establish the centrality of mental action for discussions of agency and mind. The thirteen original essays provide a wide-ranging vision of the various and nuanced philosophical issues at stake. Among the questions explored by the contributors are: Which aspects of our conscious mental lives are agential? Can mental action be reduced to and explained in terms of non-agential mental states, processes, or events? Must mental action be included among the ontological categories required for understanding and explaining the conscious mind more generally? Does mental action have implications for related topics, such as attention, self-knowledge, self-control, or the mind-body problem? By investigating the nature, scope, and explanation of mental action, the essays presented here aim to demonstrate the significance of conscious mental action for discussions of agency and mind. Mental Action and the Conscious Mind will be of interest to scholars and graduate students working in philosophy of mind, philosophy of action, and philosophy of agency, as well as to philosophically inclined cognitive scientists.
Year: 2023
Primary URL: https://www.routledge.com/Mental-Action-and-the-Conscious-Mind/Brent-Titus/p/book/9780367077518
Access Model: Subscription/ purchase
Publisher: Routledge
Type: Edited Volume
Copy sent to NEH?: No

Workshop on Addressing Ethical and Technical Challenges in the Development, Use, and Governance of Lethal Autonomous Weapons Systems (LAWS). (Conference/Institute/Seminar)
Title: Workshop on Addressing Ethical and Technical Challenges in the Development, Use, and Governance of Lethal Autonomous Weapons Systems (LAWS).
Author: Lisa Miracchi Titus
Author: Daniel E. Koditschek
Author: Michael Horowitz
Abstract: This workshop had three primary goals: Facilitate interdisciplinary communication and connections between roboticists and researchers in different fields on the subject of LAWS, to better understand differing perspectives and opinions as well as opportunities for progress. Advance our understanding of technical and ethical issues related to LAWS, and document these insights as workshop proceedings in a suitable archival venue. Create the foundation for lasting working groups to propose well-articulated positions on the regulation of autonomy in robotic weapons systems, with the long term goal of proposing to the RAS Research and Practice Ethics Committee a path toward what robotics as a discipline can contribute to the governance of AWS.
Date Range: 5/27/2022
Location: International Conference for Robotics and Automation (ICRA) 2022, Philadelphia, PA
Primary URL: https://www.grasp.upenn.edu/laws_icra2022/
Secondary URL: https://ieeetv.ieee.org/channels/techethics

A Plea for Integrated Empirical and Philosophical Research on the Impacts of Feminized AI Workers (Article)
Title: A Plea for Integrated Empirical and Philosophical Research on the Impacts of Feminized AI Workers
Author: Hannah Read
Author: Javier Gomez-Lavin
Author: Andrea Beltrama
Author: Lisa Miracchi Titus
Abstract: Feminist philosophers have long emphasized the ways in which women’s oppression takes a variety of forms depending on complex combinations of factors. These include women’s objectification, dehumanization and unjust gendered divisions of labour caused in part by sexist ideologies regarding women’s social role. This paper argues that feminized artificial intelligence (feminized AI) poses new and important challenges to these perennial feminist philosophical issues. Despite the recent surge in theoretical and empirical attention paid to the ethics of AI in general, a strikingly small portion of this work has considered the impact of feminized AI in particular. Our aim is thus to draw attention to the new and pressing ethical challenges posed by feminized AI. Doing so is crucial not just to understanding the impact of these increasingly ubiquitous technologies, but also to our understanding of longstanding feminist philosophical concerns and efforts to ameliorate them.
Year: 2022
Primary URL: https://academic.oup.com/analysis/advance-article/doi/10.1093/analys/anac050/6911414
Access Model: Subscription
Format: Journal
Periodical Title: Analysis
Publisher: Analysis

Why Intelligent Reasoning Must Be Embodied (Conference Paper/Presentation)
Title: Why Intelligent Reasoning Must Be Embodied
Author: Lisa Miracchi Titus
Abstract: This talk motivates the view that intelligent reasoning must be embodied. First it advocates for the Semantic Efficacy thesis -- the view that the contents of intelligent states and processes are causally relevant to cognition and behavior. Then it explores the implications of this thesis and motivates an anti-reductive approach to studying the foundations of intelligence. Lastly, in light of the foregoing, it discusses some recent AI systems such as GPT3 and assesses whether they shed light on the grounds of intelligence.
Date: 4/12/22
Primary URL: https://www.youtube.com/watch?v=eftmRJwkMCQ
Secondary URL: https://www.santafe.edu/events/embodied-situated-and-grounded-intelligence-implications-ai

Intelligence in Action (Conference Paper/Presentation)
Title: Intelligence in Action
Author: Lisa Miracchi Titus
Abstract: What is distinctive of real intelligence - the kind that humans and many other animals have? I’ll propose a view that is both intuitive and supported by a range of work across AI, cognitive science, and evolutionary theory. Systems with real intelligence meaningfully engage with the world. More precisely, they have semantic efficacy: the contents of their actions and mental states make a crucial difference to what they do. I’ll argue that this view is not only supported by existing empirical research but also helps us to make sense of persistent challenges in AI and cognitive science. Then I’ll explain why a commitment to semantic efficacy changes how we should do research in cognitive science and AI. Lastly, I'll discuss how this approach helps to provide clarity on some ethical issues.
Date: 2/4/22
Conference Name: University of Colorado Boulder Institute of Cognitive Science Colloquium Talk

Intelligence in Action (Conference Paper/Presentation)
Title: Intelligence in Action
Author: Lisa Miracchi Titus
Abstract: What is distinctive of real intelligence - the kind that humans and many other animals have? I’ll propose a view that is both intuitive and supported by a range of work across AI, cognitive science, and evolutionary theory. Systems with real intelligence meaningfully engage with the world. More precisely, they have semantic efficacy: the contents of their actions and mental states make a crucial difference to what they do. I’ll argue that this view is not only supported by existing empirical research but also helps us to make sense of persistent challenges in AI and cognitive science. Then I’ll explain why a commitment to semantic efficacy changes how we should do research in cognitive science and AI. Lastly, I'll discuss how this approach helps to provide clarity on some ethical issues.
Date: 2/7/22
Conference Name: University of Denver Philosophy Department Talk

Intelligence in Action (Conference Paper/Presentation)
Title: Intelligence in Action
Author: Lisa Miracchi Titus
Abstract: What is distinctive of real intelligence - the kind that humans and many other animals have? I’ll propose a view that is both intuitive and supported by a range of work across AI, cognitive science, and evolutionary theory. Systems with real intelligence meaningfully engage with the world. More precisely, they have semantic efficacy: the contents of their actions and mental states make a crucial difference to what they do. I’ll argue that this view is not only supported by existing empirical research but also helps us to make sense of persistent challenges in AI and cognitive science. Then I’ll explain why a commitment to semantic efficacy changes how we should do research in cognitive science and AI. Lastly, I'll discuss how this approach helps to provide clarity on some ethical issues.
Date: 3/17/22
Conference Name: University of Colorado Denver Philosophy Colloquium Talk

Measured interactions of task and labor allocation to feminized AIs (Conference Paper/Presentation)
Title: Measured interactions of task and labor allocation to feminized AIs
Author: Hannah Read
Author: Javier Gomez-Lavin
Author: Andrea Beltrama
Author: Lisa Miracchi Titus
Abstract: The ubiquity of gendered, particularly feminized, AIs (Alexa, Siri) raises serious ethical concerns, alongside future policy considerations, about the role that these AI assistants may have in reinforcing sexist and dehumanizing divisions of labor, particularly as many typical tasks assigned to these assistants (scheduling, reminding, caring) are thought of as stereotypically “feminine.” To date, no significant empirical interventions have been carried out on this topic. Our series of studies has begun to shed some light on this important topic by building off of related work in the domain of human-robot interaction (Kuchenbrandt et al. 2014). We hypothesized that there would be a series of at least nine interactions between AI-, participant-, and task-gender. Initial results support this complex picture. We found, for instance, that participants are more likely to assign tasks to feminized- versus masculinized-AIs (t(416) = 2.82, adjusted p < .05), and this effect is pronounced for male-participants assigned to feminized-activities (t(415) = 2.82, adjusted p < .05).
Date: 3/30/22
Conference Name: Southern Society for Philosophy and Psychology 2022 in Mobile Alabama

Permissible Uncertainty and Meaningful Human Control (Conference Paper/Presentation)
Title: Permissible Uncertainty and Meaningful Human Control
Author: Lisa Miracchi Titus
Abstract: This talk aimed to refocus the discussion of weaponized robots away from their potential physical and moral capacities, embracing instead an empirically more effective and morally more accurate focus on what information humans require to make ethical decisions about them. Humans can't be required to be infallible nor have perfect information; they cannot even require the ability to abort their robots’ commanded tasks throughout a mission since there are already many extant systems that don't carry such guarantees. Human capability to conform to the rules of engagement and LOAC can't be accomplished by merely restricting some specific scope of activity (e.g. machine may select but human must explicitly permit a lethal action) since technology is changing so rapidly. Interjecting human in loop does not necessarily increase control (e.g., if there are significant lags in communications). Introducing the notion of permissible uncertainty focuses on the appropriate epistemic conditions concerning what humans need to know or be able to think about. Traditional LOAC already accounts for an increased degree of abstraction up the chain of command which doesn't vitiate human responsibility and accountability. Permissible uncertainty for humans commanding lethal robot systems analogously requires knowing with adequate certainty that command will be accurately carried out. Several examples from the literature further motivate the appeal to permissible uncertainty by illuminating the futility of seeking specific restrictions or increased human control tied to specific levels of robot capability. Finally, this view clarifies what is needed from robotics as a discipline: designing systems (of design and development as well as testing and verification) for which legal guidance can be used to define permissible uncertainty.
Date: 5/27/22
Primary URL: https://ieeetv.ieee.org/channels/techethics/permissible-uncertainty-and-meaningful-human-control-laws-workshop-icra-2022
Conference Name: ICRA Workshop on Addressing Ethical and Technical Challenges in the Development, Use and Governance of Lethal Autonomous Weapons Systems

What Do Deep Learning Language Models Understand? (Conference Paper/Presentation)
Title: What Do Deep Learning Language Models Understand?
Author: Lisa Miracchi Titus
Abstract: Deep Learning Language Models, such as GPT3 and LaMDA, are increasingly touted as potentially shedding light on human language understanding and capacities for inference. Moreover, people both inside and outside the engineering world are raising concerns that these systems may even have become capable of understanding and communicating meaningfully themselves. In this talk, I’ll argue why we are still a long way away from developing AI systems that have these properties. Then I’ll sketch a trajectory we might take in order to close the gap between AI systems and human and animal intelligence. I’ll close by discussing some ethical ramifications of the widespread optimism about current AI capabilities and suggest a more tempered approach.
Date: 10/5/22
Conference Name: Vassar College Colloquium Talk

What Do Deep Learning Language Models Understand? (Conference Paper/Presentation)
Title: What Do Deep Learning Language Models Understand?
Author: Lisa Miracchi Titus
Abstract: Deep Learning Language Models, such as GPT3 and LaMDA, are increasingly touted as potentially shedding light on human language understanding and capacities for inference. Moreover, people both inside and outside the engineering world are raising concerns that these systems may even have become capable of understanding and communicating meaningfully themselves. In this talk, I’ll argue why we are still a long way away from developing AI systems that have these properties. Then I’ll sketch a trajectory we might take in order to close the gap between AI systems and human and animal intelligence. I’ll close by discussing some ethical ramifications of the widespread optimism about current AI capabilities and suggest a more tempered approach.
Date: 12/7/22
Conference Name: University of Glasgow, COGITO Conference

---