Fuzzy-Unweighted VBDM Framework for Value-Aligned AI Decisions
The recently introduced Fuzzy-Unweighted Value-Based Decision Making (FUW-VBDM) framework tackles the issue of ensuring that autonomous intelligent systems reflect human values. By discarding previous weights, it mitigates normative biases from stakeholders and establishes a fuzzy set of decision variables tailored for a scoring function. This innovation broadens the scope of any VBDM problem, framing it as a quest for viable solutions while optimizing the score within the weight domain. The framework integrates both qualitative and quantitative factors to produce decisions centered around human interests, with the goal of safeguarding citizen integrity and security as AI technologies become increasingly integrated into society.
Key facts
- FUW-VBDM framework introduced for value-based decision making
- Removes prior weights to address normative bias
- Uses fuzzy domain of decision variables for score function
- Generalizes VBDM as search for feasible solutions
- Incorporates quantitative and qualitative criteria
- Aims to align AI with human values
- Published on arXiv with ID 2605.13601
- Focuses on human-centered autonomous decisions
Entities
Institutions
- arXiv