AI for Not Bad

Area of Science:

• Computer Science
• Ethics in Technology
• Social Impact of AI

Background:

• The concept of "artificial intelligence (AI) for social good" is prevalent but lacks precise definition in public and data science discourse.
• Current terminology shapes public perception of AI systems, often oversimplifying their impact and development.

Purpose of the Study:

• To critically examine the vagueness and limitations of the term "AI for social good."
• To propose a more accurate and nuanced reframing for the field of AI development focused on societal impact.

Main Methods:

• Critical reflection on the conceptualization and application of "AI for social good."
• Analysis of the insufficiency of current criteria for evaluating AI's societal contributions.
• Examination of how AI systems elide externalities and structural interdependencies.

Main Results:

• The notion of "AI for social good" is characterized by vagueness and offers inadequate judgment criteria.
• The term obscures the complex externalities and structural interdependencies inherent in AI systems.
• Existing frameworks fail to capture the nuanced reality of AI's societal effects.

Conclusions:

• The field of AI development concerning societal impact should be understood as "AI for not bad" rather than "AI for social good."
• This reframing encourages a more realistic and critical approach to developing and deploying AI technologies.
• Acknowledging AI's potential for negative externalities is crucial for responsible innovation.