⚖️ Ethics · Advanced · Policy resolution

Resolved: Police in the United States should be prohibited from using facial recognition technology.

Facial recognition can help police identify suspects and find missing people, but errors and privacy concerns have led some cities to ban it. Where should the line be?

Background

Facial recognition software compares a face in a photo or video with a database of images, such as mugshots or driver’s license photos, and suggests possible matches. Police use it to generate leads in investigations. Several U.S. cities, including San Francisco (2019) and Boston (2020), banned city agencies, including police, from using it, while other places have adopted rules on how it may be used. [2]

In 2019, the National Institute of Standards and Technology (NIST) tested many algorithms on about 18 million government photos of 8.5 million people. It found demographic differences in most algorithms: in one-to-one matching, false positives for African American, Asian and Native American faces were often 10 to 100 times more frequent than for white faces, and women had more false matches than men. Results varied greatly by algorithm; some performed much better than others, and NIST said the findings weren’t definitive proof of systemic bias. Wrongful arrests after faulty matches have been reported in several cities. [1]

Where the facts themselves are disputed

Are today’s best systems accurate enough for police use?

  • Supporters say top algorithms are highly accurate and that a match is only a lead, checked by investigators.
  • Critics point to demographic differences in error rates and to wrongful arrests where matches were treated as proof.

Key terms

One-to-one matching
Checking whether two images show the same person, as when unlocking a phone.
One-to-many matching
Searching a database for anyone who matches a face, as in an investigation.
False positive
The system wrongly says two different people match.
Investigative lead
A possible match that police must confirm with other evidence before acting.

Both sides at their strongest

Each argument is written the way its best advocates would make it, followed by the strongest answer from the other side. Use them to build your case and to prepare for the other one.

Affirmative

The risks of misidentification and mass surveillance are too great; police shouldn’t use it.

  1. A1 Errors fall unevenly and can lead to wrongful arrests.

    Why If a tool is more likely to misidentify some groups, those people face a higher risk of being questioned or arrested for crimes they didn’t commit.

    Evidence NIST’s 2019 study of demographic differences; reported cases of wrongful arrests after faulty matches.

    Best answer The negative answers that the best algorithms show much smaller differences, and rules can require confirming evidence before any arrest.

  2. A2 It enables mass surveillance.

    Why Cameras plus facial recognition could track where anyone goes, chilling free movement, protest and association.

    Evidence Civil-liberties reports on surveillance networks; examples of the technology used to monitor crowds abroad.

    Best answer The negative replies that rules can limit use to specific investigations and forbid tracking crowds.

  3. A3 Rules have proved hard to enforce.

    Why Police departments have sometimes used the technology without clear policies or public knowledge, so a ban is the only reliable safeguard.

    Evidence Reports of undisclosed police use; city council hearings before bans in San Francisco and Boston.

    Best answer The negative argues that transparency laws and audits can enforce rules without banning a useful tool.

  4. A4 Communities have chosen bans.

    Why Cities that weighed the costs and benefits decided the risks outweighed the help, showing a ban is a reasonable policy.

    Evidence San Francisco’s 2019 and Boston’s 2020 bans.

    Best answer The negative replies that other cities chose regulation, and some later loosened their bans.

Negative

Used with strict rules, facial recognition helps solve crimes and find people, and shouldn’t be banned outright.

  1. N1 It helps solve serious crimes and find missing people.

    Why When investigators have only an image of a suspect or a missing child, a database search can produce leads that would otherwise take weeks or never come.

    Evidence Police reports of cases solved with facial recognition leads; use in identifying trafficking victims.

    Best answer The affirmative answers that the same leads have led to wrongful arrests.

  2. N2 Accuracy varies, and the best systems are very accurate.

    Why NIST found results differ greatly by algorithm; requiring only top-performing systems reduces errors and demographic gaps.

    Evidence NIST’s 2019 findings that algorithms vary widely and that its results weren’t definitive proof of systemic bias.

    Best answer The affirmative replies that police don’t always use the best systems, and image quality in real cases is often poor.

  3. N3 Rules can prevent misuse.

    Why Requiring a warrant or serious-crime limits, human review, and confirming evidence before arrest keeps the benefits while limiting harm.

    Evidence State and local laws regulating police facial recognition; department policies.

    Best answer The affirmative argues that rules are hard to enforce and have been broken.

  4. N4 Human eyewitnesses also make mistakes.

    Why Eyewitness identification is a leading factor in wrongful convictions; a carefully used algorithm can be one more check.

    Evidence Research on eyewitness error and exonerations.

    Best answer The affirmative replies that adding a confident-seeming computer match may make investigators trust a wrong identification more.

Strawman or steelman?

A strawman is a weak copy of the other side’s view, easy to knock down and unfair to the people who hold it. A steelman is that view at its strongest. Answer the steelman, and you’ve answered the real argument. (More on steelmanning.)

Affirmative

Strawman “Ban supporters want criminals to go free.”

Why it’s unfair Supporters worry about wrongful arrests and surveillance, which also harm innocent people.

Steelman Uneven error rates, wrongful arrests and the risk of mass surveillance are too great, and rules have proved hard to enforce, so a ban is the reliable safeguard.

Negative

Strawman “Opponents of a ban want a surveillance state.”

Why it’s unfair Most support strict rules limiting use to investigations.

Steelman With strict rules and top-performing systems, facial recognition helps solve serious crimes and find missing people; a ban throws that away.

Weak arguments to avoid

Affirmative

  • “Facial recognition never works.” Contradicted by testing; argue about error rates and misuse instead.

Negative

  • “If you have nothing to hide, you have nothing to fear.” A slogan that ignores wrongful arrests of innocent people, which is the other side’s main point.

Questions for cross-examination

Ask the Affirmative

  1. Did NIST find all algorithms equally biased?
  2. Would you allow use to find a missing child?
  3. How do eyewitness errors compare?

Ask the Negative

  1. How many wrongful arrests after faulty matches have been reported?
  2. Who checks that police use only the best algorithms?
  3. What stops use on crowds at protests?

Reading on both sides

Read at least one source from each side before you debate. Look up each source’s author and publisher too: who wrote it is part of the evidence.

Background and overviews

Both sides in one place

Mostly Affirmative

Mostly Negative

Practice this topic

Sources for the background

  1. Federal study shows face recognition accuracy varies by gender and race, Engadget (on NIST's 2019 report) (2019-12-20). Accessed 2026-10-25.
  2. Boston bans facial recognition technology, becoming another U.S. city to do so, WeLiveSecurity (2020-06-25). Accessed 2026-10-25.

Teddy Rook Book doesn’t take a side on this resolution. Updated 2026-10-25.

@teddyrookbook GitHub