How do you handle ethics concerns from users and policy enforcement in AI/ML?

Question

How do you handle ethics concerns from users and policy enforcement in AI/ML?

Asked at Google in the past year

292 views

Asked at

292 views

How to answer Technical questions

Interview Guide

Top Technical interview questions

Imagine you're the product manager for Facebook Marketplace. Since many sellers don't mark items as sold, what existing functionality and metrics could you use to determine whether an item has likely sold?7 answers | 20.9k views
What happens when you enter a URL in your browser?6 answers | 10.8k views
How does TinyURL work?5 answers | 317k views
See Technical PM Interview Questions

Invite members
Invite by email

Add another

vparam · Answer 1 · 2024-05-29T19:39:32+0000

Clarifications
- are we talking about all consumers or business or both - assumption both
- Do we provide the option to user to determine the enforcement by default or some of the option can be configurable by the users - Let us assume that enforcement does not have any user configurable options
Two major AI ethics concern regarding user policies are
- Privacy the lack of true consent to AI companies on accessing our private information and using them to enhance their models.
  - How we think on using the tools, prompts, accepting /declining of the AI provided results. We do not explicilty consent but these are used for reinforced learning by the implicit agreement to the user policies enforced.
- Bias - Models are based on the population that was used for training. While these models have improved and reduced. The AI solution user policies are enforced in a way that it does not explicit state the bias but also does not tkes the resposnibility to the bias.
Recommendation to address these ethic concerns.
- Privacy - We can be explict to the customer whether we can use their information and state though the model is proprietary and using the user information would thier contribution to having a better results though there is no compensation for the same. Alternatively we can also think about giving points or something that the user will value. But the key is to make the user enforcement being explictly accepted
- Bias - As part of enforcing the user policy on using the model, we need toe be explcit on the type of data the model is trained and continued to be fine tuned by the type of users. Again sharing the data used would be key way to address this type of ethical AI concerns while enforxing user policies.