Research

LocoLab’s research explores how local AI can support education, what small language models can actually do on consumer hardware, and how students interact with AI systems. Each paper is its own repository — published papers become public with links to the full text.

Each paper enacts one or more of the lab’s five principles. The artifacts behind these papers — built but not yet published — live in findings.

Active

These projects have scaffolded repositories and are under active development.

Paper	Description	Status	Principles
Cognitive Strategy Transfer	Framework for understanding how cognitive strategies transfer across AI-assisted learning contexts (4-paper series)	In progress	conversation; vary
DSR AI Education Simulation	Design science research on AI-powered education simulations	In progress	conversation; methodological
Keep Asking — Study 1: Does the Nudge Work?	Using frontier models, test whether a conversational nudge shifts students from passive delegation to active conversation and improves task outcomes	In progress	conversation
Keep Asking — Study 2: Does Conversation Compensate for Model Quality?	Test whether nudged students using a weak local model can match un-nudged students using a frontier model — reframing AI equity as a habits problem	Planned (pending Study 1)	conversation; vary

Planned

Early-stage ideas with initial notes. Not yet under active development.

Paper	Description	Principles
PCIe Multi-GPU Inference Scaling	Does VRAM tier or architecture generation matter more? GTX vs RTX scaling comparison on consumer hardware (experiment design)	engineer
Context Length Effects on Small Language Models	How context window size affects small language model performance on consumer hardware	specialize
Perceived Intelligence vs Token Rate	Relationship between perceived AI intelligence and token generation speed	methodological

Published

No papers published yet — check back soon.