Computer Science > Artificial Intelligence
arXiv:2512.08449 (cs)
Abstract:This paper introduces the Impact-Driven AI Framework (IDAIF), a novel architectural methodology that integrates Theory of Change (ToC) principles with modern artificial intelligence system design. As AI systems increasingly influence high-stakes domains including healthcare, finance, and public policy, the alignment problem–ensuring AI behavior corresponds with human values and intentions–has become critical. Current approaches predominantly optimize technical performance metrics while neglecting the sociotechnical dimensions of AI deployment. IDAIF addresses this gap by establishing a systematic mapping between ToC’s five-stage model (Inputs-Acti…
Computer Science > Artificial Intelligence
arXiv:2512.08449 (cs)
Abstract:This paper introduces the Impact-Driven AI Framework (IDAIF), a novel architectural methodology that integrates Theory of Change (ToC) principles with modern artificial intelligence system design. As AI systems increasingly influence high-stakes domains including healthcare, finance, and public policy, the alignment problem–ensuring AI behavior corresponds with human values and intentions–has become critical. Current approaches predominantly optimize technical performance metrics while neglecting the sociotechnical dimensions of AI deployment. IDAIF addresses this gap by establishing a systematic mapping between ToC’s five-stage model (Inputs-Activities-Outputs-Outcomes-Impact) and corresponding AI architectural layers (Data Layer-Pipeline Layer-Inference Layer-Agentic Layer-Normative Layer). Each layer incorporates rigorous theoretical foundations: multi-objective Pareto optimization for value alignment, hierarchical multi-agent orchestration for outcome achievement, causal directed acyclic graphs (DAGs) for hallucination mitigation, and adversarial debiasing with Reinforcement Learning from Human Feedback (RLHF) for fairness assurance. We provide formal mathematical formulations for each component and introduce an Assurance Layer that manages assumption failures through guardian architectures. Three case studies demonstrate IDAIF application across healthcare, cybersecurity, and software engineering domains. This framework represents a paradigm shift from model-centric to impact-centric AI development, providing engineers with concrete architectural patterns for building ethical, trustworthy, and socially beneficial AI systems.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2512.08449 [cs.AI] |
| (or arXiv:2512.08449v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2512.08449 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yong-Woon Kim [view email] [v1] Tue, 9 Dec 2025 10:21:02 UTC (333 KB)
Current browse context:
cs.AI
Change to browse by:
export BibTeX citation