“Your Brain on Software Development” by Fahran Wallace

Cognitive Biases Affect Decisions

• Systematic errors in the brain’s reasoning
• Evolutionary legacy parts for stress, predation, etc.

Brain Model

• Brain stem: automatic responses (e.g. breathing, heart)
• Limbic system: more complex stuff (e.g. emotions)
• Neocortex: Logic and thinking
• From Thinking Fast and Slow
  • Limbic & Stem: Fast (System 1)
  • Neocortex: Slow (System 2)

Fast and Slow examples

System 1, Fast

• 1+1
• Reading ‘var x = 2;’
• Recalling feelings about a particular tech
• Keyboard shortcuts System 2, Slow
• Finding security bugs
• Database isolation level
• Drawing architecture diagram

System 2 and System 1 interact

• System 1 has fast reactions or opinions to stimuli
  • what we see, hear, touch
  • recognize people, loved animals, flee from dangerous
• System 2 can override certain automatic decisions of System 1
• System 2 can train System 1 for different reactions, visual identification (looking for someone in a crowd)
• System 1 is like a hot cache of decisions
• System 2 can modify or add to the cache
• Nevertheless, System 1 is in charge by default

Cognitive Biases in Software

There’s too much information:

• anchoring effect: attachment to the first idea
  • e.g. writing mongodb as a placeholder will tend to make it permanent
  • e.g. hearing a guessed number from someone else will influence your own guess
• availability heuristic
  • i.e. when you’re holding a hammer, everything is a nail
  • i.e. if something is in System 1 cache, you might use it
  • use C4 diagrams
  • write the constraints, name no tech or product

Need to act fast:

• hard-easy effect: when a lot of pressure/rush, we do the easy things first
  • we feel there’s progress when doing the fast easy things
  • research and planning is slow, feels like no process
  • 
• Dunning-Kruger effect: confidence and knowledge about a topic are usually inversely proportional
  • despair when we turn unknown unknowns into known unknowns
  • 

    Not enough meaning:

• Patterns Where There are None
  • Correlation and causation mix-up
• Cargo Cult
• IKEA Effect: we’re biaised towards things we’ve built
• Estimation of Time and Complexity
  • we’re really bad at it
  • we can’t intuit the exponential complexity of tasks, how they interact and influence each other
  • simple tasks tend to have non-obvious edge cases

    What needs to be remembered?:

• Peak-End Effect: we remember emotional peaks (sad & happy) and the end of the project.
  • we forget the details, some lessons learned
  • use Architectural Decision Register: an immutable journal of decisions; gives the context behind decisions, explain, not repeat mistakes, not undo work, etc.

Backfire Effect

• new concepts that contradict our held worldview offend our brain
• presenting facts will backfire: it won’t convince the other person and instead might have them dig in deeper