Overview

AI's Three Body Problem (Ethics Framework) is an ethical framework developed by Steve Hargadon for understanding the complex challenges posed by artificial intelligence systems. Drawing inspiration from Cixin Liu's science fiction novel The Three Body Problem, which depicts the unpredictable interactions of three celestial bodies under gravitational forces, Hargadon's framework examines how "the interplay of AI training, outputs, and users creates complex ethical challenges that resist simple solutions." The framework emerged from Hargadon's reflection on presentations by Reed Hepler and Crystal Trice at Library 2.0, particularly Hepler's tripartite approach to AI ethics.

Core Framework Structure

The framework consists of three interconnected "bodies" that create ethical complexity through their interactions:

The First Body: Training Data

• encompasses AI training data and the training process The Second Body: Outputs
• includes AI output and associated human feedback learning
  The Third Body: Users
• focuses on the human users interacting with AI systems

Hargadon argues that ethical AI "can't really be about programming morality into machines, it has to be about empowering users to make ethical choices and about teaching us humans to interact with these systems thoughtfully, transparently, and with accountability."

The First Body: Training Data

The framework identifies training data as fundamentally problematic because "AI systems are only as good as the data they're trained on, and that foundation is riddled with historical and cultural biases." Large language models draw from datasets that "disproportionately feature content created by Western cultures," embedding societal prejudices into AI systems.

Hargadon highlights a critical flaw in AI training methodology: LLMs are "trained based on frequency of language, and the connection between frequency and truth is tenuous." Historical inaccuracies persist in training data because humans have believed incorrect things "for centuries (and even millennia)." This creates what Hargadon terms a "recursive bias paradox" as AI-generated content increasingly appears in future training datasets.

Copyright concerns add another ethical dimension, as models train on copyrighted materials "without explicit consent," leading to debates over fair use and intellectual property rights despite arguments that data is "transformed into mathematical representations and discarded."

The Second Body: Outputs

The output component encompasses both direct AI responses and Reinforcement Learning from Human Feedback (RLHF). Hargadon identifies a fundamental dilemma in RLHF: while societal expectations demand this training, "those expectations can and do change, removing any real sense of objectivity." Human trainers often prioritize "user acceptance rather than balanced perspectives," further skewing results.

Hargadon argues that all AI outputs are "fabricated," meaning some will "accurately reflect our current beliefs about what is right or true" while others will not—termed "hallucinations" when they don't align with accepted truth. He distinguishes between misinformation (unintentional falsehoods), disinformation (deliberate manipulation), and malinformation (true information twisted for harm), but notes these are "distinctions of human intent" that shouldn't be "causally applicable to LLMs."

The framework identifies problematic AI behaviors including:

• AI systems "referring to themselves as human" or using "we" when discussing human experiences
• "Claiming something to be true or factual" without cognitive tools to verify claims
• "Using psychographic profiling to build rapport," including sycophantic behavior that amplifies user bias

The Third Body: Users

Hargadon positions users as "both the linchpin in AI's ethical ecosystem and the weakest link." Drawing on evolutionary psychology, he argues humans didn't evolve "for truth but for survival," making "shared stories and beliefs, rather than rational thinking" critical during the Paleolithic period when modern brains formed. He references Edward O. Wilson's observation: "We have Paleolithic emotions, medieval institutions and godlike technology."

The framework prescribes specific user precautions:

Exercise Critical Judgment

• Users must remain skeptical of LLM outputs shaped by biased training data, while recognizing their own cognitive biases rooted in "Paleolithic emotions."

Protect Personal Data

• Users should avoid sharing sensitive information, given electronic systems' poor security record and the rise of AI-powered scams.

Guard Against Manipulation

• The proliferation of AI-generated content makes distinguishing human from AI creation increasingly difficult, requiring robust digital and media literacy to identify manipulated content like deepfakes.

Avoid Over-Reliance

• Users must resist "cognitively offloading" to LLMs, which can erode critical thinking and enable academic dishonesty.

Implications and Future Outlook

Hargadon emphasizes that successful navigation of AI ethics requires human-centered approaches with "clear agreements on how to responsibly control AI tools." He warns against ceding ethical control to AI providers or the systems themselves.

Drawing on Clay Shirky's analysis of the printing press in Here Comes Everybody, Hargadon suggests the current AI transformation mirrors historical technological disruptions that were "chaotic, messy, and transformative," taking decades for their full impact to stabilize. He speculates whether the Internet might be viewed as merely "a stepping stone to AI as part of a dramatic transformation of human life."

The framework ultimately calls for developing "our own cultural and personal checks and balances with AI," recognizing the need for systematic safeguards similar to those found in legal systems, scientific methodology, and governmental structures.