Metastimuli Learning Project
The Metastimuli Project develops methods to augment human learning by transforming conventional educational content—text, audio, or video with dialogue—into metastimuli: signals dynamically correlated with the learner’s own personal information management system (PIMS). By filtering stimuli through the graph-structured “dialectical information architecture” of a PIMS and mapping their structural position into a low-dimensional vector space, the system drives haptic actuators to create a physical, spatiotemporal accompaniment to the learning process.
Building on the Dialectical Information Architecture Project, this work integrates deep learning–based text classification, dimensionality reduction (via principal component analysis of graph metrics), and multimodal feedback design. The 2020 system introduced the concept of the metastimulus bond effect, hypothesizing that mirroring the mind’s process of categorization through correlated haptic feedback strengthens conceptual associations. The 2021 work significantly advanced the architecture through: integrated PIMS classification within the neural network, recurrent ANN variants, novel atom embedding strategies (including the ∇-embedding inspired by linguistic differentiality), keyword weighting, and meta-/hyper-parameter optimization.
The ultimate goal is a deployable platform where learners create and interact with their own PIMS, while the metastimuli system operates in real-time to enhance comprehension and retention, paving the way for rigorous experimental validation of the metastimulus bond effect.
An atom classifier derived from the atom embedder, PIMS ANN, the inverse PCA projection, and an inverse node lookup. The ANN has been trained on the user’s PIMS structure, thereby making its output a low-dimensional estimate of the atom’s “location” in the user’s PIMS. Converting this estimate into “node space” requires a nearest-neighbor search, which can occur as shown, before inverse projection, or after. (Picone et al., 2021)
Publications
- Picone, Rico A. R., Dane Webb, Finbarr Obierefu, and Jotham Lentz. "New Methods for Metastimuli: Architecture, Embeddings, and Neural Network Optimization". Augmented Cognition, Lecture Notes in Artificial Intelligence: 288–304.
- Picone, Rico A. R., Dane Webb, and Bryan Powell. "Metastimuli: An introduction to PIMS filtering". Augmented cognition. Human Cognition and Behavior 12197, Lecture Notes in Artificial Intelligence: 118–128.
Research Themes
Artificial Intelligence
Exploring the architectures and symbolic frameworks that underlie intelligent behavior in machines. This theme bridges classical and contemporary AI approaches—including logic, language models, and neural architectures—with a special focus on how AI systems represent knowledge, make decisions, and relate to human users. Psychoanalytic theory is used to interrogate assumptions about mind, subjectivity, and trust.
Augmented Cognition
Designing systems that enhance or extend human cognitive capacities through real-time feedback, machine learning, and symbolic modeling. This theme investigates how computational and robotic systems can support learning, decision-making, and self-reflection. It draws on psychoanalytic theory to understand the structural dynamics of attention, desire, and thought, and on engineering to design systems that operate in synchrony with embodied cognition.
Human-Computer and Human-Robot Interaction
Exploring the symbolic, embodied, and ethical dimensions of interaction between humans and intelligent systems. This theme examines the mutual shaping of humans and machines—how robotic and AI systems are interpreted, trusted, and engaged with by humans, and how those systems can be designed to accommodate subjectivity, ambiguity, and ethical asymmetry.
Information Architecture
Developing symbolic frameworks for the organization, representation, and transmission of knowledge. This theme includes work on dialectical systems, knowledge graphs, and semantic embeddings that formalize how meaning is structured and transformed. Psychoanalytic and linguistic concepts are brought into dialogue with computation to explore how information systems shape (and are shaped by) subjectivity.
Machine Learning
Developing learning algorithms for adaptive, nonlinear systems in uncertain and dynamic environments. This theme includes supervised, unsupervised, and reinforcement learning approaches applied to real-world systems. Particular emphasis is placed on interpretability, embodiment, and the use of machine learning to model or simulate symbolic structures, including those derived from psychoanalytic theory.
