Research Task

From application to knowledge. Each task corresponds to an unexplained phenomenon, broken down into deliverable case entries with specified evidence standards and completion criteria; items that fail review are not counted toward deliverables.

8

Research Task

5

In Progress

3

Recruiting

140 / 295

Case Submitted

Country / Region

Industry

Status

M-01In ProgressChina / United States / EuropeCross-industryDeadline2026-12

How can AI agents integrate into real organizational workflows?

Moving an agent from a demo environment into an organizational workflow typically involves three key gates: permissions, approvals, and accountability. This task tracks the real-world deployment process, documenting critical transition points from pilot to scale-up, including failures and rollbacks that occur at these junctures.

Semi-structured interviewProcess Node RecordPre- and Post-Deployment Comparison

Target Case

30row

Submitted

18row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Dr. Chen Sarah · MIT Media Lab

Submit Progress18 / 30(60%)

DeliverOrganization Flowchart and Agent Access Points

DeliverChecklist: Transition from Pilot to Scale

DeliverFailure and Rollback Case Records

View task details →
M-02In ProgressGlobal SouthManufacturing, ServicesDeadline2027-03

What determines whether AI deployment improves or replaces workers?

The same automation technology can yield opposite labor outcomes on similar production lines. This task compares four variables—skill structure, regulatory environment, organizational culture, and technical design—to distinguish the actual conditions under which automation improves jobs versus replaces them.

Cross-case comparisonJob Task BreakdownRegulatory Environment Comparison

Target Case

50row

Submitted

32row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Professor Raj Patel · University of Mumbai

Submit Progress32 / 50(64%)

DeliverFour-variable comparison table

DeliverTask Checklist Before and After Role Change

DeliverRegional Workforce Outcome Differences

View task details →
M-03RecruitingCross-regionHealthcare, FinanceDeadline2026-09

Why do identical AI systems produce different outcomes in different countries?

Institutional frameworks, languages, payment habits, and infrastructure differences can cause performance gaps for the same system across markets. This task collects deployment records of the same product or vendor in different countries to identify the sources of these disparities.

Compare with Homologous SystemInstitutional Environment CodeInfrastructure Condition Record

Target Case

40row

Submitted

12row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Dr. Maria Santos · University of São Paulo

Submit Progress12 / 40(30%)

DeliverCross-Country Performance Difference Matrix

DeliverVariance Attribution Explanation

DeliverLocalization Checklist

View task details →
M-04RecruitingEurope / AsiaGovernmentDeadline2027-06

How does AI change decision-making in government policy?

From data analysis to predictive modeling and automated decision recommendations, AI's role in public policy is progressively moving upstream. This task tracks actual usage across each stage of the decision chain and examines how accountability shifts.

Policy Process MappingDecision-making stage interviewCompare Public Files

Target Case

25row

Submitted

8row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Professor James Wright · Oxford Internet Institute

Submit Progress8 / 25(32%)

DeliverPolicy Decision Chain Diagram

DeliverAI Involvement and Responsibility Statement

DeliverPublic Case Summaries

View task details →
M-05RecruitingSoutheast Asia / AfricaAgricultureDeadline2027-09

How can agricultural AI adapt to the needs of smallholder farmers?

Agricultural AI designed for large-scale farms often fails when applied to small plots with weak connectivity and limited device budgets. This task documents the actual adoption paths of smallholder farmers, including workarounds used and reasons for abandonment.

Field Site RecordingCost Structure CalculationTrack cycle

Target Case

35row

Submitted

5row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Dr. Aisha Mohammed · University of Nairobi

Submit Progress5 / 35(14%)

DeliverSmallholder Farm Path Recording

DeliverCost and Revenue Breakdown

DeliverList of reasons for interruption

View task details →
M-06In ProgressGlobalEducationDeadline2026-11

Personalized Promise vs. Reality of AI in Education

Personalized learning claims often diverge from classroom realities. This task compares actual outcomes across different learning environments and groups, focusing on equity and accessibility rather than isolated score improvements.

Classroom ObservationLearning Process Data AnalysisTeacher-Student Interview

Target Case

45row

Submitted

22row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Dr. Yukiko Tanaka · University of Tokyo

Submit Progress22 / 45(49%)

DeliverPersonalization Mechanism Overview

DeliverEffect differences across different groups

DeliverAccessibility Issues Checklist

View task details →
M-07In ProgressNorth America / EuropeFinanceDeadline2027-01

Risk Management and Regulatory Compliance in Financial AI

Credit, anti-fraud, and market risk models improve efficiency but also introduce model risk into decision chains. This task reviews financial institutions' model governance practices and regulatory responses, documenting known failure scenarios.

Model Governance Document AnalysisCompliance Process InterviewPost-mortem of failed cases

Target Case

30row

Submitted

15row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Professor Hans Müller · ETH Zurich

Submit Progress15 / 30(50%)

DeliverModel Governance Process Overview

DeliverCompliance Requirements Comparison Table

DeliverInvalidation and Rollback Records

View task details →
M-08In ProgressChina / Germany / JapanManufacturingDeadline2026-08

Digital Transformation Roadmap for AI in Manufacturing

Quality inspection, predictive maintenance, and supply chain optimization are the first three scenarios implemented by manufacturing enterprises, but their implementation sequences and returns vary significantly. This task tracks adoption trajectories to compare how different starting points shape implementation paths.

Production Floor RecordsDevice and System Ledger VerificationPhase Return Comparison

Target Case

40row

Submitted

28row

Evidence Requirements

Each ≥ 2 lines, including ≥ 1 E1/E2

Owner

Dr. Li Wei · Tsinghua University

Submit Progress28 / 40(70%)

DeliverUse Path Timeline

DeliverBenefit comparison across three scenarios

DeliverBlocked Process Details

View task details →

Research Process

From topic to knowledge: 6 steps, each with verifiable deliverables

  1. 01

    Topic Initiation

    Identify unexplained phenomena from existing cases to formulate testable research questions, and clearly define scope and exclusions.

  2. 02

    Task Breakdown

    Break down into deliverable case entries, specifying region, industry, required evidence level, and completion criteria.

  3. 03

    Partner Claim

    Research partners claim tasks by direction. Submission requires approval: complete case coverage and traceable evidence.

  4. 04

    Evidence Verification

    Each case must include at least two pieces of evidence, with at least one being primary or authoritative secondary source; the remaining items serve as controls.

  5. 05

    Dual-track review

    The system pre-checks structural integrity and source risk; research partners verify facts and wording. Sign-off required for conclusions.

  6. 06

    Knowledge Formation

    Synthesize findings across cases, defining scope and counterexamples. Content that fails review remains hidden.

Participate in Standards

Research partner levels are assigned based on documented contributions; verbal contributions do not count. The following four criteria represent the minimum standards for admission and collaboration.

Admission
Accumulated accepted cases ≥ 10. Sign the cooperation memorandum before submission.
Contribute
Lead at least 2 knowledge atoms, each supported by at least 1 verifiable primary or authoritative secondary sources.
Review
Underwent 1 rounds of peer review; review comments are archived and accessible.
Cycle
Trial run for 7 weeks; if targets are not met, one extension of 4 week is allowed. The acceptance criteria remain unchanged.

Bring a real-world case

We prioritize cases with on-site records and traceable sources. Please ensure you can provide evidence and accept the dual-track review before submitting.

Become a Research Partner