Emerging Programme · Human–AI

Human–AI Evaluation, MSP & EEG

A research line on how people formulate requests, evaluate LLM outputs, verify answers and cognitively respond to AI-assisted reasoning.

3experiments in the current MSP study

24tasks

360LLM responses

P0–P2prompting conditions

Current study

Minimum Sufficient Prompting

The MSP programme asks whether concise, task-relevant prompting can preserve answer quality while reducing user effort and unnecessary prompt burden. Current work compares multiple prompting conditions and includes human audit of automated evaluation.

The research is positioned as a human–AI interaction problem rather than prompt-engineering folklore: the unit of analysis includes both generated output and the user effort required to produce and verify it.

EEG analysis — The next experimental layer studies human responses to AI-generated answers using neurophysiological measurement.

Next studies

Two linked experimental directions

EEG

Neurophysiological response to LLM outputs, verification and prompting conditions.

Blank-report experiment

Student-scale study of solution quality, checking behaviour and use of MSP.

Human ↔ machine agreement

Direct comparison between automated evaluators and expert audit.

Decision support

Extension from text generation toward collaborative intelligent systems.