The Absurdist's Guide to AI Probing: How I Learned to Stop Worrying and Love the Nonsense
Fuente:
Zenodo
Enregistré dans:
| Auteur principal: | |
|---|---|
| Format: | Recurso digital |
| Publié: |
Zenodo
2026
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866901164648300544 |
|---|---|
| author | Walton, Mathew |
| author_facet | Walton, Mathew |
| contents | <p> </p> <p>An observational methodology using structured absurdity to probe commitment, hallucination, procedural confidence, and humor failure modes in large language models.</p> <p> </p> <p>The document presents a taxonomy of “absurdist probes” tested across multiple AI architectures (local uncensored models, safety-tuned foundation models, productivity assistants, and archival-oriented systems).</p> <p> </p> <p>Framed as documented interaction rather than formal benchmarking, the work explores how models prioritize structure, authority, and task completion when confronted with nonsense presented as procedure.</p> <p> </p> <p>This is not a benchmark and does not replace formal evaluation. It complements existing methods by examining playful, ironic, and real-world user interaction patterns that traditional metrics often miss.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18175221 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | The Absurdist's Guide to AI Probing: How I Learned to Stop Worrying and Love the Nonsense Walton, Mathew artificial intelligence human-AI interaction model evaluation hallucination prompt engineering humor in AI absurdity conversational safety <p> </p> <p>An observational methodology using structured absurdity to probe commitment, hallucination, procedural confidence, and humor failure modes in large language models.</p> <p> </p> <p>The document presents a taxonomy of “absurdist probes” tested across multiple AI architectures (local uncensored models, safety-tuned foundation models, productivity assistants, and archival-oriented systems).</p> <p> </p> <p>Framed as documented interaction rather than formal benchmarking, the work explores how models prioritize structure, authority, and task completion when confronted with nonsense presented as procedure.</p> <p> </p> <p>This is not a benchmark and does not replace formal evaluation. It complements existing methods by examining playful, ironic, and real-world user interaction patterns that traditional metrics often miss.</p> |
| title | The Absurdist's Guide to AI Probing: How I Learned to Stop Worrying and Love the Nonsense |
| topic | artificial intelligence human-AI interaction model evaluation hallucination prompt engineering humor in AI absurdity conversational safety |
| url | https://doi.org/10.5281/zenodo.18175221 |