Why Humans Can’t Pick Truly Random Numbers

Pose the driving question and introduce a class-wide challenge: each person chooses one integer from 1 to 100 without discussion, using only a quick mental process.

• Choose one integer from 1 to 100
• Do not discuss choices before submitting
• Predict the shape of the class histogram

Learners sketch or select the histogram they expect from human choices before any results are plotted, making a testable commitment about where numbers will cluster.

• Predict the most and least common regions
• Distinguish uniformity from a human preference pattern
• Commit to one expected distribution

Collect submitted numbers into a shared live histogram so the whole class can compare its actual distribution with the earlier prediction.

• Enter each class member’s selected number
• Watch frequencies update automatically
• Mark peaks around 7, 37, 42, and similar candidates
• Compare observed frequencies with a uniform distribution

Use the histogram as evidence that human choices are structured. Introduce semantic association: 7 may feel lucky, 37 may feel distinctive, and other numbers carry personal or cultural associations.

• Meaning makes some numbers easier to retrieve from memory
• Favourite and culturally salient numbers attract attention
• Highly salient choices are repeated disproportionately
• A uniform distribution is the benchmark for equal likelihood

Manipulate two histograms and sample sizes to test whether the class pattern persists when more observations arrive or when the reference changes to a true uniform model.

• Increase the number of observations
• Compare human-choice and uniform histograms
• Test whether deviation is reduced by sample size alone
• Identify persistent peaks rather than random noise

Explain working-memory limits, avoidance of extremes, and the peak-end rule without treating any one mechanism as a complete explanation of all number choices.

• Working memory makes mentally generated sequences difficult to maintain without anchors
• People often avoid values that seem too small, too large, too round, or too obvious
• The peak-end rule helps explain why memorable episodes and salient endpoints shape remembered sequences
• The mechanisms overlap: memory supports meaning, while meaning biases selection

Test strategies in a new situation: choose an integer from 1 to 10, then compare direct choices with strategies such as cycling, anchoring on two random-looking digits, or generating from remembered lists.

• Apply the explanation to a new number range
• Compare different strategies for human randomness
• Watch strategy effects appear in the histogram
• Explain why no simple memory shortcut guarantees uniformity

Contrast human randomness with algorithmic randomness. Explain that a deterministic algorithm can generate a pseudorandom sequence whose values appear statistically random, while a CSPRNG uses entropy from the environment or another unpredictable source.

• Pseudorandom output is generated by a deterministic process
• Entropy measures uncertainty available to seed or drive a generator
• A CSPRNG is designed so predicting future outputs is computationally infeasible
• Uniformity is a property of a distribution, not proof that a generator is secure

Return to the class histogram, directly answer the driving question, and close with the central distinction: human choices carry cognitive and cultural structure, whereas engineered randomness manages entropy and statistical predictability.

• Human “random” choices are filtered by memory, meaning, and preference
• The histogram’s peaks and gaps are evidence of those filters
• Pseudorandom systems imitate statistical randomness through computation
• CSPRNGs make future outputs hard to predict using unpredictable entropy
• Your brain doesn’t generate randomness; it generates meaning