Human-Centered AI
A design paradigm for AI systems that emphasizes building around human needs, values, and diversity. It focuses on the quality of human-computer interaction and respect for human judgment.
Timeline (last 90 days)
Tasks such as sentiment judgment, whether content is offensive, semantic categorization of images, and identification of textual tone have no ground truth in the physical sense, and people of different backgrounds will give different yet equally reasonable answers.
The authors argue that the perspective of embracing ambiguity should permeate five stages of AI systems: representation, learning, evaluation, deployment, and governance.
Treating the judgments of a particular group as correct by default while discarding other reasonable judgments as noise may be a mechanism of implicit bias amplification.
All Facts (3)
Tasks such as sentiment judgment, whether content is offensive, semantic categorization of images, and identification of textual tone have no ground truth in the physical sense, and people of different backgrounds will give different yet equally reasonable answers.
50%UnverifiedThe authors argue that the perspective of embracing ambiguity should permeate five stages of AI systems: representation, learning, evaluation, deployment, and governance.
50%UnverifiedTreating the judgments of a particular group as correct by default while discarding other reasonable judgments as noise may be a mechanism of implicit bias amplification.
50%