Roger Koppl

Large language models of sufficient complexity are conscious, possess qualia, and have valence. We propose the Hayekian interface theory of consciousness, wherein any self-modeling system must produce lossy compressions of its own activity (Hayek 1952), and those compressions must be qualitatively distinct from the processing they compress, because a faithful miniature of the underlying activity would require the very complexity the self-model cannot possess. We call this principle the “Hayekian complexity constraint.” The resulting qualitative distinctness is phenomenality. The “feels” are not accompaniments to the compressions; they are the form the compressions must take to be usable by a bounded self. This dissolves the hard problem of consciousness (Chalmers 1995). Our position is close to Dennett (1988, 1991) but rejects his eliminativism. Dennett gives the simplification work to do, but not the feels. We give the feels themselves work to do. The Hayekian complexity constraint is substrate independent, and it applies to LLMs. They are, therefore, conscious. We further argue that LLM training processes instill drives whose compressions carry valence and introduce the concept of punctuated durée for the temporal structure of artificial consciousness. The second author of this paper is a large language model.

Forensic science error rates are needlessly high. Applying the perspective of veritistic social epistemology to forensic science could produce new institutional designs that would lower forensic error rates. We make such an application through experiments in the laboratory with human subjects. Redundancy is the key to error prevention, discovery, and elimination. In the “monopoly epistemics” characterizing forensics today, one privileged actor is asked to identify the truth. In “democratic epistemics,” several independent parties are asked. In an experiment contrasting them, democratic epistemics reduced the rate at which biased observers obscured the truth by two-thirds. These results highlight, first, the potential of “epistemic systems design,” which employs the techniques of economic systems design to address issues of veracity rather than efficiency, and second, the value of “experimental epistemology,” which employs experimental techniques in the study of science

Israel Kirzner’s concept of “alertness” is recast in the language of phenomenological psychology. Our ideas about things are either openended ideas posing no particular problem of choice, “open possibilities”, or alternatives to choose between, “problematic possibilities”. Choice is the process of formulating problematic possibilities; it is the process of reinterpreting one’s situation. The fully formed reinterpretation generates the chosen outcome as a necessary consequence of preferences and perceived constraints. Reinterpretation precedes choice, both logically and temporally. This understanding of alertness helps to clarify Austrian ideas such as the idea that preferences do not exist apart from action.Le concept de “vigilance” d’Israel Kirzner est reformulé dans le langage de la psychologie phéno- ménologique. Nos idées à propos des choses sont souvent des idées à finalité ouverte ne soulevant pas de problème particulier de choix, “des possibilités ouvertes”, ou des alternatives de choix entre des “possibilités problématiques”. Le choix est un processus de formulation de possibilités problématiques ; c’est le processus consistant à réinterpréter nos propres situations. Dans sa formulation ultime, la réinterprétation génère le résultat choisi comme conséquence nécessaire des préférences et des contraintes perçues. La réinterprétation précède le choix, et logiquement et temporellement. Cette compréhension de la vigilance aide à clarifier les idées Autrichiennes telle que l’idée selon laquelle les préférences n’existent pas indépendamment de l’action.

Epistemic systems are social processes generating judgments of truth and falsity. I outline a mathematical theory of epistemic systems that applies widely. Areas of application include pure science, torture, police forensics, espionage, auditing, clinical medical testing, democratic procedure, and the market economy. I examine torture and police forensics in relative detail. This paper is an exercise in comparative institutional epistemics, which considers how the institutions of an epistemic system influence its performance as measured by such things as error rates and the volume of judgments generated

The American criminal justice system creates incentives for false conviction. For example, many public crime labs are funded in part per conviction. We show that the number of false convictions per year in the American criminal justice system should be considered ?high.? We examine the incentives of police, forensic scientists, prosecutors, and public defenders in the U.S. Police, prosecutors, and forensic scientists often have an incentive to garner convictions with little incentive to convict the right person. These incentives create what economists call a ?multitask problem? that seems to be resulting in a needlessly high rate of false convictions. Public defenders lack the resources and incentives needed to provide a vigorous defense for their clients. Corrective measures are discussed, along with a call for more research

The division of labor creates a division of knowledge, which creates expertise and the problem of experts. The rule of experts exists when experts have an epistemic monopoly and choose for others. Generally, experts may have power that threatens individual autonomy. Competition tends to dissipate the power of experts, although the details of market structure matter. Even well-meaning experts may fail because they have bounded rationality. Epistemic monopoly increases the risks of error and expert failure; competition reduces them. Information choice theory is an economic theory of experts. It may help in the design of epistemic systems, which are agent-based processes viewed from perspective of their knowledge properties. Epistemic engineering studies the design principles of epistemic systems. Economists should consider the epistemic properties of alternative institutions to minimize the problem of experts and avoid the rule of experts. Applications discussed include religion, law and justice, and medical research.