Page 50 - 2622_Delhi Police Public School_C-7

Ethical Considerations in Computer Vision

Computer Vision enables machines to see and interpret the
visual world through images and videos. While powerful, it
raises following important ethical concerns:

Informed Consent

Computer Vision datasets are more often derived from the
internet or public spaces without the knowledge of individuals.
It is ethical to inform individuals that their data is being collected. However, the challenges are
identifying them and obtaining consent.

Voluntary Participation

It is ethical to collect data in a computer vision project with participants' permission. However,
content collected from social media or CCTV footage lacks this voluntary participation ethic.

Do No Harm
Computer Vision can sometimes cause problems if it makes mistakes, treats people unfairly, or is
used in the wrong way. That’s why it is important to check and test the data carefully to make sure
it is fair for everyone. This is especially important in areas like hospitals, police work, and job hiring.

Confidentiality
Many pictures and videos used to train AI are shared online without hiding people’s faces or
details. This can be unsafe. To keep data safe, it should be locked with special codes (encryption),
only shared with people who are allowed (access control), and stored in secure places where
others can’t easily get to it.

Anonymity

It’s hard to keep people completely anonymous in videos and apps that use faces. To address
this, synthetic datasets should be used, and in addition to removing names, visual data must
also be protected from identification.

Only Assess Relevant Components
Some computer systems gather more
information than they actually need. Informed Voluntary Do No Harm
Consent Participation
This is not right. Good ethics means that
systems should only collect the data that is
necessary and not go too far or try to judge Only Assess
Confidentiality Anonymity
people based on extra details. Relevant
Components

48 Premium Edition-VII