Decision Load Index: The Research
A multi-factor assessment for measuring cognitive decision burden. Open methodology. Transparent limitations. Invitation to critique.
How DLI Works
The Assessment
The Decision Load Index (DLI) measures cognitive decision burden through a multi-question assessment, producing a composite score from 0-100.
Higher scores indicate higher cognitive load (more mental burden from decisions).
What DLI Measures
The assessment evaluates decision load across several domains:
| Domain | What It Measures | Example Questions |
|---|---|---|
| Task Volume | Active commitments and open items | "How many tasks are on your to-do list?" |
| Information Backlog | Unprocessed inputs awaiting attention | "How many unread emails do you have?" |
| Context Switching | Frequency of attention shifts | "How often do you switch between projects?" |
| System Maturity | Quality of capture and review habits | "Do you have a trusted system?" |
| Subjective Load | Perceived overwhelm and decision burden | "How often do you feel overwhelmed?" |
| Cognitive Overhead | Hidden costs of task management | "How many tools do you use?" |
Scoring Methodology
The composite DLI score is calculated using a weighted algorithm:
Key weightings:
- Overdue items: Higher weight (3×) - immediate pressure
- Open loops: Moderate weight (2×) - cognitive holding costs
- Ambiguous actions: Moderate weight (1.5×) - decision friction
- Project load: Penalty for >5 active projects
Full algorithm specification available to validated research partners.
Score Interpretation
| Range | Category | Typical Context |
|---|---|---|
| 0-30 | Low (Sustainable) | Focused work, clear priorities |
| 31-60 | Moderate (Manageable) | Normal knowledge work |
| 61-80 | High (Approaching Overwhelm) | High-demand periods |
| 81-100 | Critical | System capacity exceeded |
Research Foundation
Theoretical Basis
DLI draws from three established research domains:
1. Cognitive Load Theory (Sweller, 1988)
Cognitive Load Theory distinguishes between:
- Intrinsic load: Complexity inherent to the task
- Extraneous load: Load from environment and overhead
- Germane load: Load from learning and schema construction
DLI focuses on extraneous load from decision-making overhead—the cognitive cost of managing decisions, not the decisions themselves.
2. Decision Fatigue Research (Baumeister et al., 2003)
Decision fatigue research established that:
- Decision quality degrades after repeated decisions
- Self-control and decision-making share cognitive resources
- Capacity is depletable but recoverable
DLI operationalizes decision fatigue as a measurable, trackable state.
3. NASA-TLX Inspiration
The NASA Task Load Index (Hart & Staveland, 1988) measures workload across 6 dimensions. DLI adapts key concepts for knowledge work:
| NASA-TLX Concept | DLI Adaptation |
|---|---|
| Mental Demand | Task volume + information backlog |
| Temporal Demand | Overdue items weighting |
| Effort | System maturity indicators |
| Frustration | Subjective overwhelm questions |
| Physical Demand | Not applicable to knowledge work |
| Performance | Tracked via outcomes, not self-report |
Key Difference
NASA-TLX measures post-task workload. DLI measures ongoing decision load as a leading indicator.
A Gap in AI Governance Research
Most AI security frameworks address who agents are — identity, authentication, access control. Our research addresses what they're authorized to decide, and when that authorization should be revoked automatically. The five preprints published through this initiative (DLI, CSG, NoD, Harness, Community Security) each contribute to the decision authority layer — a gap validated by our NIST NCCoE submission and acknowledged in the emerging OWASP Agentic AI Top 10 framework.
Validation & Reliability
Current Validation Status
Research Phase: Pilot (ongoing)
Status: Early validation, pre-peer-review
Validation Targets
We are designing DLI to achieve the following psychometric properties:
| Metric | Target | Status |
|---|---|---|
| Internal consistency (α) | ≥0.70 | Under validation |
| Test-retest reliability (7-day) | ≥0.70 | Under validation |
| Convergent validity (PSS-10) | r = 0.40-0.60 | Under validation |
| Convergent validity (NASA-TLX) | r = 0.50-0.70 | Under validation |
Validation study in progress. Results will be published upon completion.
Design Expectations
Based on construct analysis, we expect:
- Moderate test-retest: DLI measures state (variable), not trait (stable)
- Partial correlation with stress measures: Related but distinct constructs
- Self-report limitations: Inherent to subjective assessments
Known Limitations
- Self-report bias: All items rely on self-assessment
- Cultural variation: Initial validation with US/UK knowledge workers
- Selection bias: Pilot participants self-selected
- No gold standard: Decision load lacks an established objective measure
Transparency Principle
We document limitations because transparency builds trust.
Data Practices
What We Collect
| Data Type | Purpose | Retention |
|---|---|---|
| DLI responses | Score calculation, trend analysis | 2 years |
| Timestamps | Pattern detection (time-of-day effects) | 2 years |
| Device type | Accessibility optimization | 1 year |
| Email (optional) | Account recovery, research updates | Until deletion requested |
What We Don't Collect
- Location data
- Employer information (unless volunteered)
- Browsing history
- Third-party tracking data
Research Use
Aggregate data only. Individual responses are never shared with researchers, employers, or third parties.
Research publications use:
- Anonymized, aggregated statistics
- No individual identification possible
- Opt-in only for follow-up studies
Your Rights
- Export: Download all your data (JSON/CSV/SPSS/R formats)
- Delete: Request complete data deletion
- Opt-out: Exclude your data from research aggregation
Request data actions: research@cognitivethoughtengine.com
For Academic Researchers
Available Resources
- Methodology documentation (this page)
- De-identified aggregate datasets (upon IRB approval)
- API access for validated research protocols
- Co-authorship opportunities for significant contributions
Export Formats
| Format | Contents |
|---|---|
| CSV | Standard tabular export |
| SPSS | ZIP with CSV + syntax file + codebook |
| R | ZIP with CSV + import script + codebook |
All exports include comprehensive codebook with variable definitions.
Sample Size Calculator
For organizational pilots measuring team-level decision load:
| Desired Effect Size | Power (0.80) | Recommended n |
|---|---|---|
| Large (d=0.8) | 80% | 26 per group |
| Medium (d=0.5) | 80% | 64 per group |
| Small (d=0.2) | 80% | 394 per group |
For detecting individual change over time, minimum n=30 days recommended.
Collaboration Inquiry
We welcome:
- Validation study partnerships
- Cross-cultural adaptation research
- Intervention effectiveness studies
- Psychometric refinement proposals
Contact Research Team
Email: research@cognitivethoughtengine.com
Include: Institution affiliation, research question summary, IRB status (if applicable), timeline and resource needs
What We Don't Know Yet
Open Questions
- Optimal weighting: Are current weights optimal for prediction?
- Intervention effectiveness: Does tracking DLI improve outcomes?
- Threshold calibration: Where should "high load" warnings trigger?
- Cultural validity: Does DLI translate across work cultures?
- Objective correlates: Can we validate against physiological measures?
Planned Research
| Study | Target | Status |
|---|---|---|
| Weight optimization (regression) | Q2 2026 | In progress |
| Cortisol correlation pilot | Q3 2026 | Seeking partners |
| Cross-cultural validation (EU) | Q4 2026 | Planning |
| Intervention randomized trial | 2027 | Grant pending |
How to Contribute
If you identify methodological issues or have improvement suggestions:
Email: methodology@cognitivethoughtengine.com
We Value Critique
DLI is a research tool, not a finished product.