LearnToHire

Module 05 · Guide

Diversity, Bias & Legal Basics

Founder-led hiring is the highest-bias hiring there is: one person, moving fast, deciding on instinct. Structure is the fix — the same structure that makes hiring fairer also makes it more predictive. This is the practical version: the biases that cost you good hires, the questions you must not ask, and the UK/US rules to know.

Not legal advice. This is general guidance for founders. Employment law differs by country, state and city and changes over time — take proper advice before finalising contracts, background checks or a rejection you're unsure about.

The biases that cost you hires

Affinity bias

Rating people higher because they remind you of yourself — same university, same background, same humour.

Fix Score against written requirements, not “would I get a beer with them”. Culture add, never culture fit.

Halo / horns effect

One impressive (or bad) detail colours every other judgement — a famous employer, a typo in the CV.

Fix Score each requirement separately, before discussing the candidate overall.

Confirmation bias

Forming a view in the first two minutes and spending the rest of the interview proving it.

Fix Ask the same core questions of every candidate and record evidence verbatim before you score.

Anchoring

The first candidate you meet becomes the benchmark for everyone after them.

Fix Define what a 3 and a 5 look like before you interview anyone.

Groupthink in debriefs

The loudest or most senior voice decides, and everyone else adjusts to it.

Fix Everyone submits written scores before the debrief starts. No scores, no seat.

Pedigree bias

Over-weighting brand-name employers and elite universities as proxies for ability.

Fix Test the actual work. Pedigree predicts far less than a work sample at early stage.

Availability / recency bias

The last interview of the day feels strongest simply because it's freshest.

Fix Write scores within 15 minutes of each interview, never at the end of the day.

Non-native speaker & accent bias

Confusing fluency and polish with intelligence or capability.

Fix Assess against the communication the job actually requires — nothing more.

Designing bias out of the process

  • Write the requirements and scoring bar before you open the pipeline.
  • Ask every candidate for a role the same core questions, in the same order.
  • Use structured scorecards with evidence fields — evidence first, score second.
  • Use a work sample or paid trial task instead of relying on gut feel.
  • Get more than one interviewer on substantive stages, and collect scores independently.
  • Widen the top of the funnel deliberately: adjacent industries, career-changers, returners, non-traditional backgrounds, under-brand-named companies.
  • Strip requirements that aren't real. “Degree required” and “10 years' experience” shrink diverse pipelines for no signal gain.
  • Publish the salary range. Ranges reduce the pay gaps caused by differences in negotiation confidence.
  • Track drop-off by stage. If a group consistently exits at one stage, that stage is the problem.
  • Make interviewers accountable: no scorecard, no vote.

Job ads: what to write, what to cut

  • Describe outcomes and responsibilities, not years served.
  • List must-haves separately from nice-to-haves — and keep must-haves short.
  • State the salary range, location policy and visa/sponsorship position clearly.
  • Offer to adjust the process for accessibility needs, in the ad itself.
  • “Young, energetic team” / “digital native” — age-coded.
  • “Rockstar ninja”, aggressive superlative stacking — narrows applicant pools.
  • “Native English speaker” — say the standard of English the job needs instead.
  • Long must-have lists — under-represented candidates self-select out fastest.

Questions to avoid — and what to ask instead

The rule of thumb: ask about the requirement of the job, never the personal characteristic behind it.

Don't ask

Are you married? Do you have kids? Planning a family?

Ask instead

This role involves occasional travel to customer sites — is that something you're able to do?

Family/marital status is protected. Ask about the requirement, not the personal circumstance.

Don't ask

How old are you? What year did you graduate?

Ask instead

Tell me about the experience you have that's most relevant to this role.

Age is protected in the UK and (40+) in the US.

Don't ask

Where are you originally from? What's your nationality?

Ask instead

Are you legally authorised to work in [country], now or with sponsorship we provide?

You may confirm right to work; you may not probe national origin.

Don't ask

Do you have any disabilities or health conditions?

Ask instead

Do you need any adjustments to take part in this process? Can you perform the core duties of this role, with or without reasonable adjustments?

Pre-offer medical questioning is restricted in both the UK (Equality Act s.60) and the US (ADA).

Don't ask

What's your current salary?

Ask instead

What are your compensation expectations? Our range for this role is X–Y.

Salary-history questions are banned in many US states and perpetuate existing pay gaps.

Don't ask

What church/mosque/synagogue do you go to? Do you need Sundays off?

Ask instead

This role requires occasional weekend on-call — is that workable for you?

Religion and belief are protected characteristics.

Don't ask

Have you ever been arrested?

Ask instead

This role requires a DBS/background check at offer stage — is there anything we should discuss?

Arrest-record and 'ban the box' rules vary by jurisdiction; run checks post-offer, consistently for everyone.

Don't ask

Are you in a union? Who did you vote for?

Ask instead

Nothing — this is never relevant to a hiring decision.

Union membership and political belief attract legal protection in many jurisdictions.

UK basics

  • Equality Act 2010 defines nine protected characteristics: age, disability, gender reassignment, marriage/civil partnership, pregnancy and maternity, race, religion or belief, sex, sexual orientation.
  • Discrimination can be direct, indirect (a neutral rule with unequal effect), harassment or victimisation — indirect is where startups usually get caught.
  • Reasonable adjustments to the recruitment process are a legal duty, not a favour: extra time, written questions in advance, alternative formats, remote interviews.
  • Health and disability questions before an offer are restricted under s.60, apart from adjustments and core-function checks.
  • You must check right to work before employment starts — for everyone, not just candidates you assume may need it.
  • Positive action (encouraging under-represented applicants) is lawful; positive discrimination (choosing someone because of a protected characteristic) is not.
  • Keep interview notes: they're your evidence, and candidates can request them under UK GDPR. Write nothing you wouldn't want read back.
  • Under UK GDPR, tell candidates what data you hold, keep it only as long as you need it, and get consent before keeping them on file.

USA basics

  • Title VII, ADEA, ADA and related laws cover race, colour, religion, sex (including pregnancy, sexual orientation and gender identity), national origin, age 40+, disability and genetic information.
  • Many states and cities add protections and rules — salary-history bans, pay-range disclosure in job ads, 'ban the box' criminal-history timing.
  • ADA: no disability or medical questions pre-offer; provide reasonable accommodations during the process.
  • Form I-9 verification happens after hire; E-Verify requirements vary by state. Don't pre-screen by nationality.
  • Disparate impact matters: a neutral requirement that filters out a protected group can be unlawful even without intent.
  • Background checks under the FCRA require disclosure, written authorisation and an adverse-action process before you reject on the results.
  • Run checks and tests consistently for every candidate in the same role — inconsistency is the main evidentiary risk.

AI in hiring

  • AI screening and scoring tools are regulated in a growing number of places (for example NYC Local Law 144 bias audits, and the EU AI Act treating recruitment as high-risk).
  • Never let a model make the reject decision. Use AI to summarise and organise; keep a human accountable for every outcome.
  • Tell candidates if AI is used in assessing them, and keep the human review path open.
  • Don't paste candidate data into tools you haven't checked for data-processing terms.

Founder's five-minute checklist

  • Requirements written and scoring bar defined before interviews.
  • Same core questions for every candidate in a role.
  • Salary range published; salary history never asked.
  • Adjustments offered in writing, in the job ad and the invite.
  • Two sets of independent scores on substantive stages.
  • Background checks post-offer, consistently, with a fair process.
  • Interview notes factual, job-related and retained safely.
  • Drop-off by stage reviewed at least once a quarter.