Noise:
Chapter
10.
“Noisel
ess
Rules”
25th September
2025
Experimental
Economics
Štěpánka Slámová
Noise vs
Bias
• Bias = systematic error
• Noise = response variation
• Statistically, mostly
paying attention to bias,
but noise equally, and
often even more important
How many students are
currently enrolled at BCB?
Chapter 10:
“Noiseless
Rules”
Reducing
Noise Using
Algorithms
• Algorithm = “a process or
a set of rules to be
followed in calculations
or other problem-solving
operations, especially by
a computer” (p.123)
• Different rule-based
approaches
• Simple rules
• Improper linear models
• Machine-learning models
Improper linear
model (by Robyn
Dawes)
• An equal-weight formula
more accurate predictor
• Reason: multiple
regression computes
“optimal weights” to
minimize squared errors in
the original data, not
outside of sample
• Often better predictions
than clinical judgments
and multiple regression
Simple/Frugal
Rules
• “Combination of two or
more correlated predictors
is barely more predictive
than the best one of them
on its own” (p.127)
• → often using two is
enough
• E: bail decision
(defendant’s age + number
of past court dates
missed)
Machine
Learning
• Very large datasets, unexpected
relations
• broken-leg exceptions = special
conditions/ rare events that are
usually outside the model
Example: Bail Decisions
(Mullainathan)
• 758,027 bail decisions
• high levels of noise
• level noise (one judge
stricter in general)
• pattern noise (random
variations)
• Using ML to reduce crime
PushBack
• Professionals
generally prefer
their own judgment
• Fear of not
understanding the
algorithm
• Fear of potential
discrimination
• algorithm aversion
My life attitude after having
read “Noise”
Thanks for listening
Citations
• Album cover for The Robust Beauty [Image]. (n.d.).
Bandcamp. Retrieved
from https://f4.bcbits.com/img/a1870026119_10.jpg
• Kahneman, D., Sibony, O., & Sunstein, C. R. (2022). Noise. William
Collins.
• Kleinberg, J., Lakkaraju, H., Leskovec, J., Ludwig, J., &
Mullainathan, S. (2018). Human decisions and machine
predictions. The quarterly journal of economics, 133(1), 237-293.
• Financial Times. (2023, October 26) [Image].
Financial Times.
• Xcitium. (2025, August 6). What is an
algorithm? [Infographic]. Xcitium Cybersecurity
Blog. https://www.xcitium.com/blog/wpcontent/uploads/2025/08/what-is-algorithm.png