Sanctus, J (2025) Imperfection as a Constitutive Property of Artificial Intelligence. [Preprint]

Text AI_imperfection.pdf Download (286kB)

Abstract

Artificial intelligence (AI) is often framed as the pursuit of precision, efficiency,
and rationality. Yet human intelligence is marked by imperfection, errors, unpre-
dictability, and emotional nuance. This paper explores the paradox: as AI systems
increase in complexity, might they also begin to exhibit imperfection resem-
bling human cognition? I argue that imperfection should be reinterpreted as a
design principle rather than a flaw. Drawing on debates in bounded rational-
ity, explainability, and trust in AI Simon (1957); Mitchell (2019); Russell (2019),
I show how unpredictability and descriptive richness can serve as resources for
more socially aligned systems. Case illustrations from predictive policing and
healthcare highlight how excessive optimization can entrench bias and undermine
human judgment Lum and Isaac (2016); Topol (2019). Reframing imperfection
as a constitutive element of intelligence opens new pathways for ethical and
accountable AI. Rather than minimizing error at all costs, future AI design should
balance precision with human-like unpredictability, enabling systems that are
more adaptive, trustworthy, and socially embedded.

---

Export/Citation:

EndNote | BibTeX | Dublin Core | ASCII/Text Citation (Chicago) | HTML Citation | OpenURL

Social Networking:

Share |

---

Item Type:	Preprint
Creators:	Creators Email ORCID Sanctus, J jjesuraj37@gmail.com
Keywords:	Artificial Intelligence, Imperfection, Complexity Theory, Error, Governance, AI Ethics
Subjects:	Specific Sciences > Artificial Intelligence > AI and Ethics Specific Sciences > Computer Science Specific Sciences > Artificial Intelligence General Issues > Ethical Issues General Issues > Theory/Observation
Depositing User:	Mr J Sanctus
Date Deposited:	18 Oct 2025 12:59
Last Modified:	18 Oct 2025 12:59
Item ID:	26958
Subjects:	Specific Sciences > Artificial Intelligence > AI and Ethics Specific Sciences > Computer Science Specific Sciences > Artificial Intelligence General Issues > Ethical Issues General Issues > Theory/Observation
Date:	17 October 2025
URI:	https://philsci-archive.pitt.edu/id/eprint/26958

Monthly Views for the past 3 years

Monthly Downloads for the past 3 years

Plum Analytics

Actions (login required)

View Item