School of Law

In this paper, I consider the construction of patents as social practices. The goal is to observe patents in action, that is, to catch patents in the act of becoming patents. This method of “following the action” is well established in the sociology of science. Similar consideration of the artifices by which a new patent is staged reveals parallels to the known staging of technical papers, including the recruitment of rhetorical allies, se¬mantic fortification against subsequent challenges, and trials of cognitive strength. In each situation, assertions become stabilized facts only if subsequent recipients are in¬duced to accept them as such. However, the patent is formed in a process that largely sidesteps the mechanisms of peer review and material experimentation, substituting in¬stead legal and procedural affordances to facilitate closure. Thus, following the action from which the stabilized patent is fabricated reveals the patent as a uniquely legal, rather than technical, social object.

Artificial intelligence (AI), in the form of machine learning systems, is becoming widely deployed across many industries to facilitate the production of new technical or expressive works. Among other applications, these technologies promise rapid product design and creation, often exceeding the capacity of human creators. Commentators and policy makers have responded to these developments with a flood of literature analyzing the ways in which AI systems might challenge our existing regimes of intellectual property. But such discussions have thus far focused on entirely the wrong questions, misunderstanding the nature of the changes that AI brings to creative development. Intellectual property is generally styled as a solution to the “appropriability” or “public goods” problem in creative and innovative production: offering a legally enhanced incentive to invest in goods that are expensive to produce, but cheap to appropriate. But cost savings from AI systems will largely occur at a different point in the production process. AI systems promise (or threaten) to lower the cost of initial development of creative goods, potentially displacing human creators. Although machine learning systems are realistically unlikely ever to provide a complete substitute for human creative inputs, their incorporation into creative production will in effect automate the generative phases of the creative development process, substantially lowering the cost of the initial stage of production. Like other cost-saving industrial automation, this can be expected to displace human labor and redefine human roles in production. The history of past automated labor displacements teaches us something of what will occur as creativity is automated. In this light, I begin to reframe the discussion of intellectual property and artificial intelligence, showing the impact machine learning will have on human creativity and innovation, and the implications these changes for intellectual property doctrine and policy. In particular, I show that cheap substitutes for human creativity will drive a shift toward forms of intellectual property that certify authenticity rather than those that incentivize production and distribution. Armed with this understanding, we can begin to address the question of how to foster human engagement in an age of synthetic creativity.

Forensic feature-comparison examiners compare-or "match"-evidence samples (e.g., fingerprints) to provide judgments about the source of the evidence. Research demonstrates that examiners in select disciplines possess expertise in this task by outperforming novices-yet the psychological mechanisms underpinning this expertise are unclear. This article investigates one implicated mechanism: statistical learning, the ability to learn how often things occur in the environment. This ability is likely important in forensic decision-making as samples sharing rarer statistical information are more likely to come from the same source than those sharing more common information. We investigated 46 fingerprint examiners' and 52 novices' statistical learning of fingerprint categories and application of this knowledge in a source-likelihood judgment task. Participants completed four measures of their statistical learning (frequency discrimination judgments, bounded and unbounded frequency estimates, and source-likelihood judgments) before and after familiarization to the "ground-truth" category frequencies. Compared to novices, fingerprint examiners had superior domain-specific statistical learning across all measures-both before and after familiarization. This suggests that fingerprint expertise facilitates domain-specific statistical learning-something that has important theoretical and applied implications for the development of training programs and statistical databases in forensic science. (PsycInfo Database Record (c) 2023 APA, all rights reserved).

The misinformation effect occurs when people's memory of an event is altered by subsequent inaccurate information. No study has systematically tested theories about the dynamics of human hippocampal representations during the three stages of misinformation-induced false memory. This study replicates behavioral results of the misinformation effect, and investigates the cross-stage pattern similarity in the hippocampus and cortex using functional magnetic resonance imaging. Results show item-specific hippocampal pattern similarity between original-event and post-event stages. During the memory-test stage, hippocampal representations of original information are weakened for true memory, whereas hippocampal representations of misinformation compete with original information to create false memory. When false memory occurs, this conflict is resolved by the lateral prefrontal cortex. Individuals' memory traces of post-event information in the hippocampus predict false memory, whereas original information in the lateral parietal cortex predicts true memory. These findings support the multiple-trace model, and emphasize the reconstructive nature of human memory.