How Can Algorithms Improve Hiring Decisions?

TL;DR

Hiring decisions become easier when you separate sampling, workload testing, and switching into structured processes. Use the 37 rule to establish a benchmark before selecting, apply TCP-style workload increases to discover capacity safely, and use the Gittins index principle to recognize when exploring a less-tested alternative may offer more value than continuing with an inconsistent option.

Transcript

i recently read algorithms to live by by brian christian and tom griffiths one day you'll need to hire someone if you haven't done so already you may hire a freelancer online to help you with your next business project you may hire a new team member at work or you may hire someone to clean your house or look after your kids for a few hours a week r... Read More

Key Insights

• The 37 rule is a selection method that uses the early part of a search to establish a benchmark, then recommends choosing the next available option that exceeds that benchmark. It helps balance the risks of committing too quickly and searching for too long.
• The 37 rule works best when rejected options cannot be revisited. When earlier candidates are likely to remain available, the transcript describes a modified approach that allows a longer sampling period and permits returning to the strongest person evaluated if no later candidate proves better.
• A benchmark is valuable because preferences often become clearer through comparison. Evaluating candidates without immediately hiring gives the decision-maker practical evidence about quality, priorities, and tradeoffs, making later choices more disciplined than relying on an untested initial impression.
• TCP-inspired onboarding is a capacity-testing process that starts with a small task and expands workload after each successful result. This approach lets a manager discover a new hire's capabilities quickly while limiting the damage caused by assigning too much work before reliability is established.
• Additive increase multiplicative decrease responds to overload by cutting the workload sharply, then rebuilding it gradually. Applied to onboarding, the method gives a struggling worker room to recover, strengthen confidence, and approach the previous failure point with better preparation and a more manageable progression.
• The Gittins index is presented as a method for deciding whether to continue with a proven option or explore an alternative. It compares evidence from successes and failures while also recognizing that uncertainty can make a less-tested project or person worth evaluating.
• Exploration has value because testing a new option produces information that continued reliance on a familiar option cannot provide. A newcomer with limited evidence may deserve consideration even when early results appear weaker, because uncertainty leaves greater room for discovering superior long-term potential.
• Structured algorithms reduce decision stress by turning vague judgments into repeatable procedures. Selection, onboarding, and switching become separate problems with distinct rules: establish a benchmark before choosing, increase responsibility according to demonstrated capacity, and explore alternatives when current performance is insufficiently reliable.

Questions & Answers

Q: How can the 37 rule improve hiring decisions?

The 37 rule improves hiring by separating evaluation from selection. During the opening portion of the candidate pool, the employer interviews people only to understand the available quality and create a benchmark. After that sampling period, the employer selects the next candidate who is better than everyone already interviewed. The rule offers the best chance described in the transcript, but it does not guarantee the desired outcome.

Q: When should the 37 rule be used for candidate selection?

The 37 rule is most appropriate when a decision must be made sequentially and previously rejected options cannot be recovered. That condition commonly appears when candidates may move on or lose interest after an employer passes. If earlier candidates are likely to remain available, the transcript recommends modifying the strategy by sampling longer and potentially returning to the strongest earlier candidate when later interviews produce no improvement.

Q: Why should employers sample candidates before hiring?

Sampling candidates before hiring helps the employer discover what strong performance actually looks like. The initial interviews reveal differences in skills, fit, and quality, which creates a practical benchmark for later comparisons. Without sampling, an employer may commit before understanding the candidate pool. With excessive sampling, the best available person may become unavailable, so the algorithm seeks a balance between those risks.

Q: How can TCP principles help onboard a new hire?

TCP principles can guide onboarding by treating workload as a capacity test. Give the new hire a small assignment first, then increase the workload after work is returned on time and according to expectations. Continue expanding responsibility while performance remains successful. When the workload causes failure or unacceptable work, reduce it sharply so the person can recover before responsibilities begin increasing again.

Q: What should a manager do when a new hire becomes overloaded?

A manager should reduce the workload substantially when a new hire becomes overloaded, following the additive increase multiplicative decrease pattern described in the transcript. The reduction creates room to catch up and restore reliable performance. After stability returns, the manager should add work gradually instead of immediately restoring the failed workload, allowing skills and confidence to develop before the person reaches that demanding level again.

Q: How does the Gittins index guide switching decisions?

The Gittins index guides switching by comparing the success and failure records of available options while accounting for uncertainty. The option with the stronger index is treated as the better choice to pursue. In practical terms, the framework helps decide whether to keep investing in a current employee or project, or test an alternative that may have similar potential but less evidence available.

Q: Why can a less-tested option be more valuable than a familiar one?

A less-tested option can be more valuable because uncertainty creates an opportunity to discover better performance. A familiar option may have a long record showing only average reliability, while a newcomer still has unresolved potential. Trying the newcomer also produces information, which improves future choices. The transcript therefore treats exploration as valuable, especially when the established option is not consistently meeting expectations.

Q: When should an employer consider replacing a freelancer or new hire?

An employer should consider switching when the current person repeatedly fails to meet expectations and another candidate appears to have comparable potential. The transcript emphasizes tracking successes and failures instead of relying only on impressions. Replacement is unnecessary when performance is consistently satisfactory, but inconsistent results create a reason to explore. Testing a new option may reveal stronger performance and provides useful information even when the outcome remains uncertain.

Summary & Key Takeaways

• The selection algorithm addresses the danger of deciding either too early or too late. The 37 rule recommends using the opening portion of a search only to evaluate options and establish a benchmark. After that sampling period, choose the next available candidate who performs better than everyone previously evaluated.

• The onboarding algorithm adapts ideas from transmission control protocol and additive increase multiplicative decrease. Begin with a small assignment, increase the workload after successful delivery, and cut it sharply after failure. Once the person recovers, increase work gradually so capacity develops without repeatedly overwhelming the new hire.

• The switching algorithm uses the Gittins index principle to balance perseverance with exploration. A familiar option offers evidence, while a newer option offers information and potential improvement. When a current hire or project performs inconsistently and another option has comparable potential, testing the less-known alternative can create valuable evidence for future decisions.