Mental Models and Heuristics

Thinking tools to solve thorny problems. Software engineering isn’t just about production, it’s about learning in complex, uncertain situations. This talk explores how useful (if imperfect) models and practical heuristics can help you reframe problems, cut through complexity, and make better technical decisions.

Transcript

1. Mental Models And Heuristics photo: Sebastian Schuster
2. What's a Mental Model? A mental model is a simplified representation of reality. It helps us explain and communicate how something works by highlighting key information, while ignoring irrelevant details. Confoo 2026 - Marc-Antoine Aubé
3. "All models are wrong, but some are useful". — George Box Confoo 2026 - Marc-Antoine Aubé
4. What's a Heuristic? noun a rule or piece of information used in or enabling problem-solving or decision-making. Confoo 2026 - Marc-Antoine Aubé
5. Building a Repertoire of Thinking Tools photo: Dan Crile
6. Type 1 vs Type 2 Decisions photo: Benjamin Lehman
7. Heuristic #1 Don't apply Type 1 decision making process to Type 2 decisions. Confoo 2026 - Marc-Antoine Aubé
8. Sudoku vs Minesweeper photo: John Morgan
9. Sudoku Problems — What you see is what you get — No way to get more clues — You move forward by making educated guesses — You end up exploring and testing a lot of dead ends Confoo 2026 - Marc-Antoine Aubé
10. Minesweeper Problems — You can easily and cheaply gather more information — You move forward by probing, observing and collecting — Given enough clues, the right answer is obvious Confoo 2026 - Marc-Antoine Aubé
11. Heuristic #2 Never Sudoku a problem when you can Minesweeper it. — Simon Tatham Confoo 2026 - Marc-Antoine Aubé
12. Open-Loop and Closed-Loop Systems photo: Aleksandr Popov
13. Open-loop System — Simple but "dumb" — Operates without feedback — Does not monitor or adjust its output based on the results photo: Daniel Salgado
14. Closed-loop System — Complex but "smart" — Goal seeking — Integrates feedback to achieve desired goal photo: Dan LeFebvre
15. When Open-Loop Goes Wrong — No monitoring — No early warning alerts — No automatic cleanup — The only "feedback loop" was us getting paged at 2am Confoo 2026 - Marc-Antoine Aubé
16. Heuristic #3 Ask: What would break if I went on vacation for a while? Confoo 2026 - Marc-Antoine Aubé
17. Mechanical Sympathy "You don't have to be an engineer to be be a racing driver, but you do have to have Mechanical Sympathy." — Jackie Stewart, racing driver Confoo 2026 - Marc-Antoine Aubé
18. The 1 Billion Row Challenge Calculate the Min, Max, and Average of 1 Billion Measurements
19. Heuristic #4 Know your system's theoretical ceiling. Confoo 2026 - Marc-Antoine Aubé
20. Theory of Constraints There is always one constraint in the system that limits flow. Call that constraint a bottleneck. The only way to improve flow is to identify and manage the constraint. Improvements made elsewhere are meaningless. All that matters is managing the constraint. — Dr Eliyahu Goldratt Confoo 2026 - Marc-Antoine Aubé
21. Heuristic #5 Work on improving the current constraint to flow. Confoo 2026 - Marc-Antoine Aubé
22. 3X Thinking Explore, Expand and Extract Confoo 2026 - Marc-Antoine Aubé
23. Heuristic #6 Explore: reduce the cost of experiment and invest in many unrelated experiments. Expand: ease the limitations of the next rate-limiting resource. Extract: find economies of scale, standardize, increase revenue and decrease costs. Confoo 2026 - Marc-Antoine Aubé
24. Chesterton's Fence photo: Clay LeConey
25. If you don't know what problem it was solving, you can't know if that problem still exists. Confoo 2026 - Marc-Antoine Aubé
26. Real Example Removed: "Useless" database index Result: Critical job started timing out 3 weeks later Why it existed: 2am data volume spikes The problem: Context was lost—no comment, no ticket reference Confoo 2026 - Marc-Antoine Aubé
27. Heuristic #7 Before removing or "fixing" something that seems pointless, ask: What problem was this solving? Confoo 2026 - Marc-Antoine Aubé
28. The Sociotechnical System Programming properly should be regarded as an activity by which the programmers form or achieve a certain kind of insight, a theory, of the matters at hand. — Peter Naur, Programming as Theory Building, 1985 Confoo 2026 - Marc-Antoine Aubé
29. Symmathesy noun A symmathesy (sim-MATH-uh-see) is a learning system made of learning parts. So!ware teams are each a symmathesy, composed of the people on the team, the running so!ware, and all their tools. Coined by Nora Bateson in 2015
30. What This Looks Like In Practice Ship → usage spikes Break → search times out at 3am Fix + Measure → add cache + metric Learn → users prefer fuzzy matching Adapt → add query suggestions New bottleneck → autocomplete service The system learned. The people learned. The so!ware changed. Confoo 2026 - Marc-Antoine Aubé
31. Heuristic #8 A!er solving a problem: What did we just learn, and how do we keep it? Confoo 2026 - Marc-Antoine Aubé
32. Eight Different Lenses About making decisions About understanding systems About knowing when to change your approach Heuristics need to be challenged. What works at 10 engineers might not work at 100. What works in a startup might not work in a regulated industry. Confoo 2026 - Marc-Antoine Aubé
33. Build Your Own Toolbox Pick one or two that resonated. But more importantly: start capturing your own. Confoo 2026 - Marc-Antoine Aubé
34. How to Start — Notice questions you already ask — Distill what you hear at conferences — Index cards, running docs, Slack channels The format doesn't matter. Capturing it does. Confoo 2026 - Marc-Antoine Aubé
35. Share What You Learn Code reviews: Share the model you're using Design docs: Make assumptions explicit Retrospectives: Capture the lesson, not just the action The value isn't having the right answer. It's building a shared vocabulary for thinking about problems. Confoo 2026 - Marc-Antoine Aubé
36. Build Your Mental Toolbox Confoo 2026 - Marc-Antoine Aubé
37. Thank You! Confoo 2026 - Marc-Antoine Aubé