AI and Ethics Computer science

Misconceptions of artificial intelligence: the modern seer as a synthesis of human knowledge without consciousness, emotion, or moral sill

Sarfaraz K. Niazi, Mikołaj Stańczak
Springer Nature · 2026-11-20 · DOI 10.1007/s43681-026-01527-z

Abstract

Abstract Artificial intelligence is often depicted in public discourse as either a compliant statistical instrument or an emerging entity endowed with knowledge, creativity, feelings, and intentions that may soon rival human capabilities. Both portrayals are fundamentally flawed, as they rely solely on observable performance and infer internal experiences from social appearances. This Viewpoint advocates for a more rigorous model: conceptualizing artificial intelligence as a modern seer, an intricate system that compresses, retrieves, and recombines recorded human knowledge at scales unattainable by any individual. The Viewpoint develops the seer from a metaphor into an analytical framework. The seer is a consulted but uncertified epistemic authority: its capabilities generate interaction cues, the cues invite inferences of understanding, emotion, intention, and authority, and governance must act on these cues and on the calibration of trust rather than on claims about a machine mind. This model clarifies both the practical utility of these systems and the misguided authority often attributed to them. It demonstrates that synthesis can generate novel insights without consciousness, that affective language can provide reassurance without genuine concern, and that value-sensitive recommendations can be produced without moral agency. These distinctions do not render the technology inherently safe, as systems lacking experience can still amplify errors, entrench biases, concentrate power, and undertake consequential actions. The field of drug development exemplifies these limitations: a model may integrate evidence throughout the pharmaceutical process; nevertheless, biological validation and accountable judgment remain essential. This discussion is no longer solely philosophical; regulatory bodies such as the United States Food and Drug Administration, the European Medicines Agency, the International Committee of Medical Journal Editors, and prominent scientific publishers have each adopted frameworks to evaluate context, risk, credibility, and transparency, none of which directly address machine consciousness. Institutions have already moved beyond metaphor-based policymaking. This Viewpoint encourages scientists and scholars to do the same. It identifies ten misconceptions, eight category errors and two factual errors, traces each to the interaction cue that produces it, and states what the framework adds to existing debates on anthropomorphism, stochastic parrots, responsibility gaps, and context-based governance. Addressing these misconceptions will change how AI systems are tested, trusted, and governed.

Read at publisher →