Robots in Disguise

Abstract

This paper argues that generative artificial intelligence (GAI), designed to mimic human engagement through natural language understanding (NLU), obfuscates the distinction between human and machine in ways that misguide rational users — particularly when addressing sensitive questions requiring accuracy, expertise, and empathy. Through a comparative case study of a user researching a Schizophrenia diagnosis across Google search and tiered ChatGPT models (Legacy GPT-3.5, Default GPT-3.5, and GPT-4), the paper demonstrates how paid-tier systems amplify access inequality, obfuscate the reliability and origin of information, and fail to meet normative standards of ethical behavior. The paper calls for GAI organizations to implement transparent codes of conduct: disclosing limitations in greater detail, disclosing inequality perpetuated by paid-tier systems, directing users to credible organizations and professionals for support, and adopting an ethical decision-making model informed by consequentialism and a rights-and-duties approach. Note: This paper shares a case study and empirical data with its companion piece, "AI's Influence on Socially Constructed Kinds" (2023), which grounds the same observations in the philosophy of interactive kinds and looping effects. The two papers are intended as complementary treatments: the companion provides the descriptive and analytical framework, while this paper presents the normative and prescriptive argument.

Ähnliche Arbeiten