The Ethics of Facial Recognition Technology

Privacy and Security and Artificial Intelligence

Article Snapshot

Author(s)

Brenda Leong and Evan Selinger

Source

The Oxford Handbook of Digital Ethics, Carissa Véliz, ed., forthcoming, 2021

Summary

Facial recognition technology presents a unique threat to freedom and human dignity. Once the technology is installed, people will become acclimated to it, and abusive uses will be nearly inevitable.

Policy Relevance

Policymakers should limit the use of facial recognition technology before the infrastructure is installed.

Main Points

  • Facial recognition systems come in four main types:
     
    • Facial detection applications find human faces.
       
    • Facial characterization systems collect information such as gender, approximate age, and emotional indicators.
       
    • Verification systems use biometrics to distinguish one human being from another.
       
    • Identification systems search for a match between one unknown face (perhaps from a surveillance camera) and many known faces (such as a database of driver’s license holders).
       
  • Testing shows that the best systems are over 99 percent accurate, but error rates among other systems are high, especially when analyzing the faces of people of color, transgender people, and women; the risk of harm from misidentification is not evenly distributed.
     
  • Humans also have high error rates when identifying faces.
     
  • Perfect surveillance can be harmful; individuals under constant surveillance might avoid wrongful conduct because they are aware that others are watching, not because they are capable of making moral decisions.
     
  • The key question is, can any level of facial recognition be implemented that will not lead to a society of ubiquitous public surveillance?
     
  • Some "slippery slope" arguments present wild speculations, but others point to specific and plausible mechanisms of decline; for example, once video cameras are installed, the costs of adding intrusive surveillance capabilities becomes much lower.
     
  • Facial recognition systems present uniquely dangerous slippery slope risks.
     
    • Faces play a special role in human life.
       
    • Faces reveal a great deal of information compared with other biometrics.
       
    • The costs of extracting information from faces are low compared with DNA or fingerprints, because a vast store of facial images already exists.
       
    • There are few legal limits on the use of facial recognition technology.
       

Get The Article

Find the full article online

Search for Full Article

Share