Earlier applied research from Cognitive Thought Engine. DLI is not part of the core Enterprise Agent Architecture runtime-governance stack — see cognitivethoughtengine.com for CTE's current work.
An author-developed 10-question cognitive load assessment for AI-augmented workplaces.
The Decision Load Index measures cognitive decision load — the cumulative weight of open loops, unresolved commitments, and system fragmentation that impairs a knowledge worker's ability to make clear, confident decisions.
DLI is not a stress scale or a burnout inventory. It measures the structural conditions that produce cognitive overload: task accumulation, context-switching frequency, inbox backlog, tool fragmentation, and the presence or absence of trusted capture and review systems.
It was developed and studied at Cognitive Thought Engine LLC in the context of AI-augmented work, where the proliferation of AI tools often increases, rather than reduces, decision burden.
- Not a clinical diagnostic instrument
- Not a replacement for occupational health assessment
- Not validated on clinical or medical populations
pip install dli-instrumentRequires Python 3.10+. No runtime dependencies — pure standard library.
from dli import calculate_dli_score, QUESTIONS
# Print all questions and options
for i, q in enumerate(QUESTIONS, 1):
print(f"Q{i}. {q.text}")
for j, opt in enumerate(q.options):
print(f" {j}. {opt}")
# Score responses: each integer is the 0-based index of the selected option
# 0 = first option, 4 = last option, in Q1-Q10 order
responses = [2, 3, 2, 1, 2, 2, 2, 2, 2, 2]
result = calculate_dli_score(responses)
print(result.summary())
# DLI Score: 19/40
# Band: HIGH LOAD
# Percentile: 57th (relative to the 901-user study sample)
#
# Elevated decision load. Consider simplifying commitments or improving capture systems.
# Check if the score is elevated
if result.is_elevated:
print("Score is in HIGH LOAD or above.")
# Per-question breakdown
for qs in result.per_question_breakdown:
print(f"{qs.question_id}: score={qs.score}, selected='{qs.selected_option}'")# Interactive assessment
dli assess
# Score pre-collected responses (10 integers, 0-4)
dli score 2 3 2 1 2 2 2 2 2 2
# Show band thresholds and normative data
dli infoimport csv
from dli import calculate_dli_score
with open("participants.csv") as f:
reader = csv.DictReader(f)
for row in reader:
responses = [int(row[f"q{i}"]) for i in range(1, 11)]
result = calculate_dli_score(responses)
print(f"{row['id']}: {result.score} ({result.band}, {result.percentile:.0f}th pct)")Each question has 5 options (indexed 0-4). Pass responses as a list of 10 integers to calculate_dli_score().
| Question | Scoring | Rationale |
|---|---|---|
| Q1 — Tasks on to-do list | Forward (0-4) | More tasks = more load |
| Q2 — Project switches/day | Forward (0-4) | More switching = more load |
| Q3 — Unread emails | Forward (0-4) | More backlog = more load |
| Q4 — Trusted capture system | Forward (0-4)* | Having a system reduces load |
| Q5 — Weekly review frequency | Forward (0-4)* | Regular review reduces load |
| Q6 — Browser tabs open | Forward (0-4) | More tabs = more load |
| Q7 — Feeling overwhelmed | Forward (0-4) | More overwhelm = more load |
| Q8 — Productivity apps used | Forward (0-4) | Tool fragmentation = more load |
| Q9 — Know what to work on | Forward (0-4)* | Clarity reduces load |
| Q10 — Work on unimportant tasks | Forward (0-4) | Priority drift = more load |
Total score range: 0-40
* Q4, Q5, and Q9 measure a protective construct (presence of the thing reduces load), but no index reversal is applied at scoring time — their answer options are pre-ordered so index 0 already means minimum load and index 4 already means maximum load, same as every other question. reverse_scored is retained on each QuestionScore as metadata identifying which items measure a protective factor, not as a signal that scoring transforms the index.
| Score | Band | Interpretation |
|---|---|---|
| 0-8 | LOW LOAD | Lower observed decision load relative to the study sample. |
| 9-16 | MODERATE LOAD | Manageable load. Monitor for accumulation over time. |
| 17-24 | HIGH LOAD | Elevated decision load. Consider simplifying commitments or improving capture systems. |
| 25-32 | VERY HIGH LOAD | Very elevated decision load. Review workload structure and open commitments. |
| 33-40 | CRITICAL LOAD | Extremely elevated decision load. Prioritize reducing task accumulation and fragmentation. |
Score a completed assessment.
Parameters
responses: List of exactly 10 integers, each 0-4, in Q1-Q10 order.
Returns DLIResult dataclass with:
score: int— total score (0-40)band: str— band labelinterpretation: str— human-readable interpretationpercentile: float— percentile rank vs. the study sampleper_question_breakdown: list[QuestionScore]— per-question detailis_elevated: bool— True if score >= 17 (HIGH LOAD or above)is_at_risk: bool— deprecated alias foris_elevated, kept for backward compatibilitysummary() -> str— formatted summary string
Return (band_label, interpretation) for a score.
Return interpretation string for a score.
Estimate percentile rank relative to the 901-user study sample using the standard normal CDF approximation (mean=18.2, std=7.4).
The DLI was studied in a self-selected sample of 901 technology-adjacent knowledge workers (online cohort, predominantly US-based, working in AI-augmented professional environments).
| Statistic | Value |
|---|---|
| n | 901 |
| Mean | 18.2 |
| Std | 7.4 |
| Distribution | Approximately normal |
| Population | Knowledge workers, AI-augmented workplaces |
| Collection period | 2025-2026 |
Limitations: The study sample is not population-representative. It skews toward technology-adjacent workers who self-selected into an online cognitive load assessment. Use comparisons against this sample cautiously in dissimilar populations.
The hosted web version of this assessment has been retired. The instrument itself is unchanged and fully usable via this package — pip install dli-instrument and score yourself with the functions documented above.
Saleme, M. (2025). Decision Load Index: A cognitive load diagnostic for AI-augmented workplaces. Zenodo. https://doi.org/10.5281/zenodo.18217577
@misc{saleme2025dli,
author = {Saleme, Michael},
title = {{Decision Load Index: A Cognitive Load Diagnostic for AI-Augmented Workplaces}},
year = {2025},
publisher = {Zenodo},
doi = {10.5281/zenodo.18217577},
url = {https://doi.org/10.5281/zenodo.18217577},
note = {Studied in a self-selected sample of 901 technology-adjacent knowledge workers. Software: https://github.com/CognitiveThoughtEngine/dli-instrument}
}@software{saleme2025dli_software,
author = {Saleme, Michael},
title = {{dli-instrument: Decision Load Index Python package}},
year = {2025},
publisher = {GitHub},
url = {https://github.com/CognitiveThoughtEngine/dli-instrument},
note = {pip install dli-instrument}
}# Clone and install in editable mode with dev dependencies
git clone https://github.com/CognitiveThoughtEngine/dli-instrument.git
cd dli-instrument
pip install -e ".[dev]"
# Run tests
pytest tests/ -v
# Run example
python examples/basic_usage.pyMIT License. See LICENSE.
Copyright (c) 2025 Cognitive Thought Engine LLC
Cognitive Thought Engine LLC researches the intersection of AI augmentation and human cognitive performance. Our work focuses on how AI tools reshape decision-making load, organizational knowledge structures, and knowledge worker wellbeing.
- Research: doi.org/10.5281/zenodo.18217577
- Current work: cognitivethoughtengine.com
- GitHub: github.com/CognitiveThoughtEngine