I am conducting a study to evaluate Google's responsible AI review process.
Context: When AI solutions and technologies are proposed, the Google Responsible AI team conducts a review to ensure the proposed solution/technology aligns with its established AI principles.
I would like to apply Rate (1994) Ethical Decision Making (EDM) framework to validate the integrity of the review process, and to establish if an EDM framework is useful in augmenting the level of confidence in the intended outcomes of the proposed AI solutions
I would welcome thoughts, suggestions, and ideas on an evaluation approach
Thank you
The challenge with AI technologies / solutions is their inherent bias, as subconscious bias in the way that AI is structured, influences the results and impacts differently upon different communities. To understand the different ways that AI impacts are experienced, the review process should ideally be able to represent the differing impacts that occur for communities having different values and beliefs, as it is what is important to different communities that determines how they experience an outcome. Therefore the different prioritising of outcomes by different communities needs to be understood, quantified and incorporated into an assessment of the potential positive and negative impacts of AI.
This is certainly an interesting study. However, I was wondering that your study may have subjectivity in exploring Google’s ethical standard on the basis of its evaluation principles disclosed in the survey after the technology or@ AI is built. It’s because ethical issues are integrated into the codes/parameters and the algorithm built as well as how the algorithm was trained with what quality of data. unless you examine them thoroughly it will be hard to understand the reality as to how algorithm would behave ethically. Ethics are subjectively constructed and interpreted, and algorithm may interpret data (Fx. Word or name representing a gender) from one culture in a particular way while it may interpret the similar data from another culture quite differently. Anyway, I Wish you best of luck with your study.
Mohammad Bakhtiar Rana thanks for your thoughts. The plan is the evaluate the review process of the proposed AI solution. this review process determines pursue vs. Not pursue. Welcome further thoughts. Thanks
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence