scieee AI-readable full text Open interactive document viewer

Reasoning with Small Language Models (SLM) to Create Trustworthy GenAI

Clark, Jason A.

Publisher
Zenodo
DOI
10.5281/zenodo.17644430
Repository
zenodo.org
Language
en

Abstract

A central problem with current interfaces for GenAI are the rapid modes of output production. These implementations are optimized for efficiency and concealing the ways output is created. I would suggest that there’s no malice here, but rather, an human-computer interaction (HCI) oversight that misses how to create systems we can trust and learn from. We trust what we can see and explain. With the rise of inference models and chain of thought coding techniques, there are new ways to slow down these systems and ask them to show their work processes to create content. In this talk, I’ll walk through a working LLM agent prototype and emerging research using small language models (models that run on a phone or laptop), applied reasoning/inference methods, and an interface that works to explain the process it used to come up with an answer. In the end, we’ll come to understand how we can take advantage of the affordances of the dialogic interface and how we can start to build trust in these systems. I’ll also make the case for how experiments like this are essential components of GenAI Literacy and how you can bring these lessons into your own work to understand and teach GenAI.

Full text

Jason A. Clark 2025 Reasoning with Small Language Models (SLM) to Create Trustworthy GenAI Jason A. Clark @jasonclark Montana State University DLF 2025 Jason A. Clark 2025 What’s the problem? Jason A. Clark 2025 Jason A. Clark 2025 Problem Statement November 30, 2022 We were given access to a weirdly confident intern that can’t explain how they got their answers. It is a seamless interface that breaks search interaction + retrieval patterns. Jason A. Clark 2025 Jason A. Clark 2025 Problem Statement More crawling, fewer referrals. More telling, less linking. It is a seamless interface that breaks search interaction + retrieval patterns. Source: “The crawl before the fall… of referrals” Cloudflare Radar Jason A. Clark 2025 What can we do? Jason A. Clark 2025 Jason A. Clark 2025 Research Motivation - Speculative Design, Frictional AI, Seamful Design Rethink the human-computer interaction. An artifactual interface that shows its work? Jason A. Clark 2025 Jason A. Clark 2025 Prototype - “Agentic Search” Artifact using Small Language Models Jason A. Clark 2025 Jason A. Clark 2025 Prototype - “Agentic Search” Artifact Agent Label + “Thinking” Observation to Focus Context Jason A. Clark 2025 Jason A. Clark 2025 Prototype - “Agentic Search” Artifact Clear Action Final Answer with Sources Trust Signal Jason A. Clark 2025 Jason A. Clark 2025 Push our values into these systems. (Claude Integration) Model Context Protocol Server and Configuration Integration with Consumer LLM Jason A. Clark 2025 Jason A. Clark 2025 Push our values into these systems. (Claude Integration) Model Context Protocol Server and Configuration Trust Signal + Sources in Consumer LLM Jason A. Clark 2025 Jason A. Clark 2025 Research Implication Language Models + Trust - Try and practice through implementation. - Natural Language Interfaces + language models for retrieval are here. - Be the best Hater (or Stan) you can be. Jason A. Clark 2025 **Thank you** Jason A. Clark @jasonclark (Bluesky + GitHub) Jason A. Clark 2025 Jason A. Clark 2025 Language Models - Tools and Getting Started ●Context Engineering ○System Prompt Example ○Deep Background - Mike Caulfield ●Running Local Language Models ○Ollama ○LM Studio ●Example Code ○@jasonclark GitHub - Gists Jason A. Clark 2025 Jason A. Clark 2025 References and Follow-up Resources Dunne, A., & Raby, F. (2013). Speculative everything: Design, fiction, and social dreaming. The MIT Press. https://mitpress.mit.edu/9780262019842/speculative-everything/ Ericson, P., Khairova, N., & De Vos, M. (Eds.). (2024). HHAI-WS 2024: Proceedings of the Workshops at the Third International Conference on Hybrid Human-Artificial Intelligence (co-located with HHAI 2024), Malmö, Sweden, June 10–11, 2024. https://ceur-ws.org/Vol-3825/ (Frictional AI Workshop) Natali, C., Naiseh, M., Cabitza, F., & Frischmann, B. (2025). Better AI with designed friction: Theories, applications and research agenda. In Frontiers in Artificial Intelligence and Applications: Vol. 408: HHAI 2025 (pp. 518–521). IOS Press. https://doi.org/10.3233/FAIA250680 Upol Ehsan, Q. Vera Liao, Samir Passi, Mark O. Riedl, and Hal Daumé III. (2024.) Seamful XAI: Operationalizing Seamful Design in Explainable AI. Proc. ACM Hum.-Comput. Interact. 8, CSCW1, Article 119 (April 2024), 29 pages. https://doi.org/10.1145/3637396