My research is organized around a single theoretical arc that moves between the philosophy of memory and artificial intelligence. It begins with a proceduralist account of semantic memory, which challenges the long-standing assumption that knowledge is stored and retrieved from a mental repository. Instead, I argue that semantic memory is best understood as a constructive, embodied capacity. This account provides the conceptual basis for my second trajectory: the application of procedural memory frameworks to AI, where it informs questions of representation, architecture, and alignment. Taken together, these projects aim to reshape how we think about what memory is and what kinds of systems—human or artificial—can have it.

Proceduralism about Semantic Memory

My dissertation builds on research in philosophy and cognitive science to develop a new account of semantic memory. While episodic memory (memories of past personal experiences) has been the central focus of philosophical inquiry, semantic memory (memory for facts) has often been treated as theoretically unproblematic: a storehouse of factual knowledge, language use, and general information about the world. But the conceptual underpinnings of this “orthodox view” are far less stable than often assumed. It presupposes a dedicated storage space of explicit facts, shaped by unconscious mechanisms of abstraction and generalization. Yet this view faces persistent challenges from both philosophical and empirical work: the lack of evidence for stable, dedicated semantic storage; difficulties locating or characterizing the neuroanatomical substrates of semantic memory; and the rise of enactivist approaches that reject stored content altogether.

In response, I propose a proceduralist alternative. On this view, semantic memory is not stored content but the procedural skill to reconstruct semantic knowledge in context. Rather than treating semantic recall as retrieval from a mental database, we can understand it as the skilled enactment of mnemonic abilities shaped by prior experience and social practice. This view integrates multi-trace memory frameworks with enactivist cognition, and draws on Wittgenstein’s notion of ungrounded hinges, Moyal-Sharrock’s non-cognitive certainties, and Rowlands’ concept of Rilkean memory.

Several short-term publications flow directly out of the dissertation, each developing and refining its core ideas:

• The Nature of Semantic Memory and the Orthodox View - critiques the standard view of semantic memory as a storage space for factual knowledge and develops a procedural alternative grounded in multi-trace and enactive approaches. Semantic memory is not stored content but the capacity to reconstruct knowledge through skilled, embodied activity.

• Mnemonic Effort and Mnemonic Habit - argues that semantic memory operates through the interaction of deliberate mnemonic effort and embodied mnemonic habit. Efforts involve the active reconstruction of knowledge from prior experience, while habits provide a stable, non-propositional background akin to Wittgenstein’s hinges or Rilkean memory.

• The Phenomenology of Semantic Recall - examines the phenomenological structure of semantic remembering, focusing on the interplay between noetic consciousness (the felt sense of knowing) and anoetic consciousness (the embodied, habitual dimension). It argues that semantic recall is structured by the mutual reinforcement of these two modes, much of which is experientially liminal—present in experience without being directly accessible to reflective attention.

• Memory and the Metaphor of Origami - proposes a shift away from the “storage and search” metaphor toward an origami metaphor for memory. Memory traces are reconceived not as stored representations but as procedural folding instructions, emphasizing memory’s constructive and reconstructive nature. This reframing clarifies what is preserved in memory and how it is reactivated.

My long-term goal is to develop these ideas into a book-length treatment of semantic memory, offering both a critical examination of the dominant view and a positive account of memory as a dynamic, constructive capacity. This work advances philosophical debates in memory theory while bridging them with cognitive science and philosophy of mind.

Artificial Intelligence and Procedural Knowledge

The second trajectory of my research grows organically out of the first. If semantic memory is not a storehouse of facts but a constructive, procedural capacity, then this has direct implications for how we design and understand intelligent systems. This trajectory unfolds along three interrelated lines of inquiry:

• AI and the Orthodox View of Memory - Large language models (LLMs) generate complex outputs without explicitly storing facts. Their performance reveals the limitations of the storage metaphor: generative capacity can emerge from procedural prediction and reconstruction rather than explicit content retrieval. This parallel challenges classical assumptions in both cognitive science and AI, and positions proceduralist accounts of memory as a conceptual bridge between the two domains. I use LLMs as a philosophical lens for rethinking the cognitive underpinnings of semantic memory itself.

• Procedural Memory Architectures for AI - Building on this insight, I explore how proceduralist memory might inform the architecture of artificial intelligence systems. Current AI designs often mirror the orthodox model of semantic memory: they treat knowledge as a static representational database. I argue that human-like intelligence may require systems that can forget, transform, and reconstruct representations flexibly, much like human mnemonic processes. This project forms the basis of a paper on procedural knowledge representation, contrasting symbolic and connectionist models with dynamic, enactive architectures.

• Virtue Cultivation and AI Alignment - The third branch extends these ideas to the alignment problem in AI. Drawing on virtue ethics and enactivism, I argue that role-playing games and immersive virtual environments can function as virtue cultivation mechanisms, scaffolding moral dispositions in AI systems in ways analogous to human enculturation. Rather than encoding narrow rule sets, this approach emphasizes procedural virtue alignment—the development of flexible, context-sensitive ethical capacities through interaction.