Human-AI Interaction

Scope

Study AI-assisted writing, decision support, retrieval, multimodal exchanges, and automation. Separate interface presentation from backend state, and human approval from a generated assertion that approval occurred.

Questions worth testing

• What context was visible to each participant?
• Which revision or branch led to the final object?
• What supports a claim of human review or approval?

Evidence and study material

Conversation exports, source databases, revision history, retrieved documents, tool records, and authorized interface captures can preserve different aspects of the exchange. Consent and collection scope matter when records contain other people’s material.

Methods to begin with

• acquisition
• reconstruction
• reporting

Limits on interpretation

A prompt can be quoted, injected, exploratory, or later rejected. The same text may have different roles in different interfaces. Avoid assigning motive from isolated messages or treating a system response as a witness account.

Open research questions

Questions include automation effects, authenticity judgments, and how source labels persist through revisions. Human-subject research needs an appropriate consent and ethics process separate from technical artifact collection.

Tools, projects, datasets, and publication

The interaction guide and SFIF provide a starting vocabulary; Synthetic Imprinting offers an adjacent cognitive perspective. This site hosts no participant dataset. Journal work should disclose study design, permissions, and material AI involvement.

• Tools and documentation resources
• Current project status
• Available reference data and dataset limits
• Journal

What information was available during the interaction?

Distinguish displayed conversation history from the context supplied to a model step. Examine available instructions, retrieved excerpts, uploads, summaries, and truncation records. Treat a generated summary as a derivative with its own limitations.

Explain which parts of the context are directly recorded. A later explanation by the user or model can be relevant context without replacing the missing historical input or proving the reason for a particular action.

Related fields & glossary terms

• Interaction Forensics
• AI Agent Forensics

Apply the examination methods, inspect the evidence classes, and check relevant external standards and specifications.