Human-Centered AI Augmentation

Definition

Human-centered AI augmentation is the design and governance stance that AI should increase people’s agency, learning, creativity, health, and problem-solving capacity while leaving human dignity, judgment, motivation, and responsibility intact.

Current Synthesis

The Fei-Fei Li episode makes augmentation both a technical and civic principle. AI can help people write, communicate, learn, discover, generate images or video, interpret health information, train robots, and imagine new environments, but those abilities should serve human purposes rather than become a reason for experts or companies to decide society’s path alone.

A complementary scientific-workflow account treats language models as idea generators, search and comparison aids, prediction tools, or clinical partners. Their advantage can differ from human expertise—for example, broader exposure to rare cases—so combined performance may exceed either alone. Augmentation still requires experts to test hypotheses, evaluate evidence, recognize context, and remain accountable for consequential decisions, as illustrated in How to Improve at Learning Using Neuroscience & AI | Dr. Terry Sejnowski.

Key Claims

• AI should expand human agency and dignity rather than remove motivation, authorship, responsibility, or public voice.
• Augmentation depends on the user’s or institution’s purpose: a model is more useful when people know what they want to learn, make, diagnose, or govern.
• High-stakes domains need expert judgment and social rules around AI, not only more capable models.
• Teachers, parents, students, creators, clinicians, and the public are not downstream users only; they are stakeholders in whether AI use remains human-centered.
• The augmentation frame rejects both doom-only and utopian narratives because both can crowd out concrete human problems and institutional responsibilities.
• Complementarity can arise when models and experts have different error patterns, but combined performance is not automatic.

Evidence

• Agency evidence: Using AI to Increase Your Intelligence & Enrich Humanity | Dr. Fei-Fei Li records Li arguing that AI should support human agency, motivation, and dignity rather than take them away.
• Domain evidence: Using AI to Increase Your Intelligence & Enrich Humanity | Dr. Fei-Fei Li applies augmentation to writing, healthcare, science, education, robotics, and creative work rather than treating it as a narrow productivity claim.
• Governance evidence: Using AI to Increase Your Intelligence & Enrich Humanity | Dr. Fei-Fei Li criticizes public AI rhetoric that talks down to people or leaves decisions to a small industry group.
• Education evidence: Using AI to Increase Your Intelligence & Enrich Humanity | Dr. Fei-Fei Li says students should not be denied useful tools, but AI should not remove their agency or motivation to learn.
• Creative evidence: Using AI to Increase Your Intelligence & Enrich Humanity | Dr. Fei-Fei Li presents AI video and World Labs as creator-supporting tools whose value still depends on human storytelling, emotion, and technique.
• Research evidence: How to Improve at Learning Using Neuroscience & AI | Dr. Terry Sejnowski presents LLMs as “idea pumps” that can suggest experiments and help compare literature, with proposals still requiring testing.
• Clinical complementarity: How to Improve at Learning Using Neuroscience & AI | Dr. Terry Sejnowski reports a dermatology example in which doctors and AI perform better together, while the exact figures remain source-scoped.

Counterevidence & Qualifications

These sources provide a human-centered argument, not a guarantee that every AI deployment will preserve agency or improve performance. Augmentation can fail when tools become answer machines, errors correlate rather than complement one another, automation bias overrides expert dissent, institutions use AI to avoid responsibility, or private companies define public norms without participation. The dermatology percentages and research-workflow examples are episode-reported and lack full methods in the supplied notes.

What Changed

• Added complementary-error and expert-verification requirements from Sejnowski’s research and medicine examples.
• Extended the concept from agency-centered design to hypothesis generation and evidence synthesis.

Related Concepts

• AI Assistant Augmentation - everyday task-assistant version of the same human-led tool stance.
• Human Agency Under AI - broader agency question that augmentation is meant to protect.
• Human Judgment Under AI - responsibility layer needed when augmented work still affects real people.
• Human-Centered AI Education - education-system version of human-centered augmentation.
• Medical AI Robot Collaboration Boundary - medical and surgical version where augmentation must remain clinician-led.
• AI Creative Collaboration - creative-work version where AI broadens options but humans retain authorship and taste.
• AI Governance And Compliance - institutional rule layer needed when AI affects public rights and safety.