SL-004Stage 01: Concept and scopingPlanned field study

Mentored team learning in shared AI research workspaces

Research questionHow do shared AI workspaces affect teaching, feedback, coordination, and accountability in research teams?

Review the design View source context

SL-004 public status

Current stage: Concept and scoping
Recruitment: Not started
Results: None
Last reviewed: 28 August 2026

PLANNED STUDY CASE FILE

A Learning Question With a Testable Decision

This case file examines the research group as a learning environment. A shared AI workspace may expose intermediate reasoning, preserve feedback, and help mentors see where learners need support. It may also increase surveillance, blur responsibility, or encourage premature delegation. The proposed study would combine workflow measures with direct evidence of learning and clearly separate adoption from effectiveness.

Learning context: Laboratory groups, methods teams, capstone research courses, or supervised research programs where teaching and production occur in the same workflow.
Decision this study should support: Determine whether a shared workspace should be studied in a controlled multisite design and identify the governance conditions required for responsible instructional use.

Protocol objectives

01
Describe how feedback, revision, and escalation patterns change after workspace adoption.
02
Estimate changes in research-method knowledge and independent task performance.
03
Measure whether decisions and AI contributions become more traceable.
04
Identify harms involving privacy, surveillance, deference, workload, and responsibility ambiguity.

DESIGN CANDIDATE

What Would Be Tested

Every element remains provisional until the protocol is registered. Unknowns are shown as unknowns instead of being filled with unsupported precision.

Design candidate: Prospective mixed-method field study with baseline, adoption, and follow-up periods
Unit of analysis: Research team, with nested learner and task observations
Comparison: Interrupted time series, staggered adoption, or matched comparison teams considered during design
Data sources: Minimized event logs, artifact histories, surveys, interviews, observation, and independent learning tasks
Sample size: Not set. The number of teams and repeated observations must support the intended team-level inference
Field duration: Not set. A baseline and follow-up long enough to distinguish novelty from stable practice are required

Learning sequence

Model

Mentors make expert reasoning, quality criteria, and escalation decisions visible in the shared record.

Practice

Learners complete bounded research tasks with explicit responsibility and allowed AI assistance.

Review

Teams annotate decisions, identify errors, compare alternatives, and document feedback and revision.

Transfer

Learners perform an independent task so workspace activity is not mistaken for durable learning.

MEASUREMENT

Outcomes Defined Before Observation

Candidate roles may change during protocol review. Any primary outcome will be fixed before data collection or access to relevant outcome data.

Outcome	Role	Operational definition	Timing
Feedback latency	Candidate workflow outcome	Elapsed active working time between a review request and substantive mentor feedback under prespecified rules.	Baseline and adoption periods
Decision traceability	Candidate workflow outcome	Proportion of sampled material decisions linked to evidence, responsible person, rationale, and revision history.	Repeated artifact audit
Independent learning performance	Candidate learning outcome	Score on a task completed without shared-workspace assistance using a domain-relevant rubric.	Baseline and follow-up
Appropriate escalation	Safety outcome	Rate at which learners seek qualified review for prespecified high-risk or low-confidence decisions.	Throughout adoption
Perceived surveillance and workload	Implementation and harm outcome	Validated or cognitively tested survey items supplemented by confidential interviews.	Baseline, adoption, and follow-up

ANALYSIS DISCIPLINE

A Plan That Can Report Unfavorable Results

Final estimands, models, exclusions, missing-data rules, multiplicity decisions, and stopping conditions will be specified in the registered protocol where applicable.

01
Specify the team-level causal or descriptive estimand before selecting the final comparison design.
02
Model time, team clustering, repeated learners, and adoption timing where the design supports these terms.
03
Triangulate event logs with artifact audits and interviews instead of treating platform activity as learning.
04
Report implementation variation and negative cases, including teams that reduce or stop use.
05
Avoid causal language if adoption is self-selected and no defensible comparison is available.

RESEARCH INTEGRITY

Implementation, Access, and AI Disclosure

The record must be detailed enough to audit what learners experienced, what the AI system could do, and where qualified humans remained responsible.

Implementation record

Workspace configuration, permissions, retention, model details, integrations, and policy changes
Team roles, mentoring structure, research-task categories, and baseline workflow
Core instructional practices and local adaptations
Event definitions, logging coverage, missing periods, and participant-visible data dictionary
Incidents, opt-outs, access changes, and deviations from agreed governance

Equity and access

Assess whether access, language, disability, time zone, employment status, or seniority changes participation in the shared record.
Provide private channels for feedback that should not be visible to the full research team.
Separate mentoring evaluation from employment or academic progression decisions unless explicitly governed.
Document who can view, export, correct, and delete workspace data within institutional constraints.

AI-system disclosure

Workspace vendor, plan, model options, version dates, retention settings, and administrator controls
Integrations, files, external sources, tools, and permissions available to the AI system
Visibility of prompts and outputs across team roles
Rules for confidential, unpublished, controlled, or personally identifiable information
Human approval responsibilities for scientific, teaching, authorship, and governance decisions

VALIDITY REGISTER

Main Risks and Planned Responses

These responses reduce specific risks. They do not eliminate uncertainty or guarantee that the final design will support a causal claim.

R01

Adoption bias

Record reasons for participation, compare baseline characteristics, and limit causal interpretation when adoption is not assigned.

R02

Novelty effects

Include a sufficient baseline and follow-up, model time explicitly, and record changes in enthusiasm and burden.

R03

Surveillance and power

Use data minimization, role-based access, private reporting routes, and independent consent procedures.

R04

Activity mistaken for learning

Include independent performance tasks and qualitative evidence of reasoning, not only usage counts.

R05

Responsibility diffusion

Assign accountable humans for material decisions and audit whether escalation rules are followed.

STUDY GATES

Status Changes Require an Exit Record

A stage label is a public claim. The record moves forward only when its stated exit condition is documented.

Concept

Current

Target teams, intended inference, governance constraints, and feasible comparison options are recorded.

Protocol drafting

Not started

Data minimization, outcomes, comparison, field duration, and analysis plan are fixed.

Review

Not started

Scientific, ethics, employment, privacy, security, accessibility, and institutional reviews are documented.

Registration and execution

Not started

Protocol and governance materials are time-stamped before observation begins.

Reporting

Not started

Benefits, harms, implementation variation, uncertainty, and case reports are released.

EVIDENCE CONTEXT

Real Sources That Inform the Protocol

These external sources provide context for design decisions. They are not Santaros Labs outputs, endorsements, or evidence that this proposed study has been completed.

National Institute of Standards and Technology · 2023 and current resources

Artificial Intelligence Risk Management Framework

Informs explicit governance, measurement, management, documentation, and human accountability for AI-system risks.

Open source

UNESCO · 2023

Guidance for generative AI in education and research

Frames human-centered use, privacy, institutional validation, teacher capacity, learner agency, and inclusive access.

Open source

Institute of Education Sciences · Living standard

Standards for Excellence in Education Research

Supports relevant outcomes, implementation evidence, generalizability, open science, and reporting that includes null and negative findings.

Open source

OPEN SCIENCE PLAN

Planned Public Open

Field-study protocol
Workspace governance template
Participant-facing data dictionary
Event-log specification
Mentoring observation guide
Independent learning-task rubric
De-identified implementation case reports

Availability will depend on consent, ethics review, licensing, privacy, security, and institutional requirements. A restriction will be explained rather than presented as open access.

Next portfolio record

SL-007AI literacy and teacher judgment in Lithuanian classrooms

METHODS COLLABORATION

Improve This Study Before Registration

We welcome educators, learning scientists, domain researchers, statisticians, research software specialists, and governance reviewers who can strengthen the protocol.

Start a research inquiry

A research inquiry does not imply study enrollment, institutional approval, funding, or authorship.