📖 Book Club
Intermediate
~2h

How to Think About AI

Master Richard Susskind's Critical Frameworks

Master Richard Susskind's critical frameworks for AI evaluation. Essential thinking tools for the AI era. For a complementary perspective, see our AI Critical Thinking course.

By Richard Susskind • Published 2024 • Critical Thinking Foundation

TL;DR:

Think critically about AI using Susskind's frameworks: shift from "is AI intelligent?" to "what can AI do?", balance process vs. outcome thinking, evaluate across seven risk categories, and plan for multiple futures—move from hype to structured analysis.

Guide byAlbert Schaper
How to Think About AI — 5 frameworks: Intelligence→Capability, Process vs Outcome, Transformation Types, AI Futures Matrix, Risk Categories

Intelligence-to-Capability Shift

Judge AI by what it does (capability), not whether it "thinks" like humans (intelligence).

Intelligence-to-Capability Shift

❌ Traditional Focus:

Focus on consciousness, understanding, human-like reasoning

✅ New Thinking:

Focus on outcomes, performance, practical utility

Process vs. Outcome Thinking

The most powerful framework: Distinguish between trusting the process vs. trusting the outcome.

Process vs. Outcome Thinking

🔄 Process-Thinking:

  • • Trust doctors because they went to medical school
  • • Trust judges because they follow proper procedures
  • • Worried AI won't replace them because "process is wrong"

🎯 Outcome-Thinking:

  • • Want the diagnosis that's correct
  • • Want legal research that finds relevant cases
  • • AI can deliver outcomes even without human-like process

Automation, Innovation, Elimination

Three ways AI transforms work—each requires different strategy.

Automation, Innovation, Elimination

Automation

AI does existing tasks faster/cheaper

Example: Automated email responses, data entry

Innovation

AI enables new capabilities previously impossible

Example: Personalized learning at scale, drug discovery

Elimination

AI makes entire processes obsolete

Example: Traditional homework, manual translation

Five AI Futures Matrix

Five possible scenarios for AI development—plan for multiple futures, not just one.

Five AI Futures Matrix
FutureLikelihoodTimelineStrategy
HypeLow3-5 yearsDon't over-invest; question assumptions
GenAI+HIGHOngoingContinuous adaptation; invest strategically
AGIMedium-High2030-35Plan long-term; develop resilience
SuperintelligenceMedium2040+Existential questions matter
SingularityLow-Medium???Emphasize adaptability

Seven Risk Categories

Systematic framework for AI risks—move from vague anxiety to structured thinking.

Seven Risk Categories
Technical Risks: Errors, hallucinations, brittleness
Security Risks: Adversarial attacks, data breaches
Bias & Fairness: Discriminatory outcomes, representation gaps
Privacy: Data collection, surveillance, consent
Economic: Job displacement, inequality, market concentration
Social: Misinformation, manipulation, trust erosion
Existential: Misalignment, loss of control, catastrophic outcomes

AI Tool Evaluation Framework

Apply Susskind's frameworks when evaluating any AI tool:

✅ Questions to Ask:

  • Capability: What can this tool actually do?
  • Outcome: Does it solve my problem effectively?
  • Process: Can I understand how it works?
  • Type: Is this automation, innovation, or elimination?
  • Risks: Which of the 7 categories apply?
  • Future: How does this fit multiple AI scenarios?

❌ Avoid These Traps:

  • ❌ Asking "Is it intelligent?" instead of "What can it do?"
  • ❌ Only process-thinking OR only outcome-thinking
  • ❌ Treating all AI as just "automation"
  • ❌ Vague anxiety instead of structured risk analysis
  • ❌ Predicting one future instead of planning for multiple
  • ❌ Evaluating based on hype instead of frameworks

Frequently Asked Questions

What is the intelligence-to-capability shift?▾
Stop asking "Is AI intelligent?" and start asking "What can AI do?" Judge AI by outcomes and capabilities, not by whether it thinks like humans. This shift is essential for practical AI evaluation.
Why is process vs. outcome thinking so important?▾
Process-thinking trusts the method (e.g., "doctors went to medical school"). Outcome-thinking trusts the result (e.g., "diagnosis is correct"). AI can deliver outcomes without human-like processes. Evaluate tools from BOTH perspectives.
How do I identify automation vs. innovation vs. elimination?▾
Automation: AI does existing tasks faster. Innovation: AI enables new capabilities. Elimination: AI makes processes obsolete. Each requires different strategy—automation optimizes, innovation explores, elimination transforms.
Which AI future should I plan for?▾
Susskind recommends planning as if AGI is likely (most challenging scenario) while staying responsive to GenAI+ reality (current state). Don't predict one future—prepare for multiple scenarios with adaptive strategies.
How do I systematically evaluate AI risks?▾
Use the seven categories: Technical, Security, Bias, Privacy, Economic, Social, Existential. Evaluate each AI tool across ALL categories, not just one. This moves from vague anxiety to structured thinking.

Apply It: Evaluate an AI Tool with Susskind's Frameworks

  1. 1Choose an AI tool you use or are considering. Browse AI Tools
  2. 2Capability Shift: What can this tool actually DO? List 3 concrete capabilities — not marketing claims.
  3. 3Process vs. Outcome: Can you understand HOW it reaches its outputs? Or do you only see the result?
  4. 4Three-Way Classification: Is this tool automating a task, innovating a new approach, or eliminating a role entirely?
  5. 5Risk Assessment: Apply at least 3 of Susskind's 7 risk categories to this tool (bias, privacy, accountability, etc.).
Reflect: Did the framework change your opinion of the tool? Which risk category concerned you most? How would you explain your evaluation to a colleague?

Apply Susskind's Frameworks

AI Critical Thinking
Deepen your ability to separate hype from reality
Browse AI Tools
Evaluate real tools with your new frameworks
AI Fundamentals
Understand the technology behind the capabilities you evaluate
Related Courses & Guides
Superagency: What Could Go Right with AI
📖 Book Club
Intermediate
New Laws of Robotics: Defending Human Expertise
📖 Book Club
Intermediate
The Coming Wave: Technology & Containment
📖 Book Club
Intermediate–Advanced
Book Club · Our take 2026
Various (curated frameworks) · 2024

Editorial assessment: How to Think About AI

Strong complement to our Critical Thinking course — best for readers who want philosophical and policy lenses, not tool tutorials.

Strengths

  • Multiple perspectives on risk and opportunity
  • Good book-club discussion fuel

Limitations

  • Not a hands-on manual
  • Overlap with other critical-thinking content on the site

Discussion questions

  1. Which framework in the book changed your mind the most?
  2. Where do ethics and business incentives conflict in your industry?
  3. What would you measure to know AI is helping, not just speeding up?

Next on Academy: AI Critical Thinking

Key Insights: What You've Learned

1

Think critically about AI using Susskind's frameworks: shift from "is AI intelligent?" to "what can AI do?", balance process vs. outcome thinking, evaluate across seven risk categories, and plan for multiple futures—move from hype to structured analysis.

2

Apply systematic evaluation by considering capability (what AI can do), risk (what could go wrong), governance (how to manage it), and futures (multiple possible outcomes)—structured thinking prevents both over-optimism and excessive fear.

3

Master AI evaluation by using these frameworks consistently: assess capabilities realistically, identify risks systematically, design governance thoughtfully, and plan for uncertainty—critical thinking about AI enables better decisions, better tools, and better outcomes.

Test Your Knowledge

Complete this quiz to test your understanding of Susskind's critical AI frameworks.

Loading quiz...

Risks & Responsible Use

Know these before you go further.

Anecdote ≠ Evidence

Narrative books persuade with stories; they are not peer-reviewed capability benchmarks.

What this means for you

Extract frameworks, then validate claims against primary sources and your own pilots.

Publication Lag

Books lag model releases by 12–24 months — tactics may predate agents, RAG, and new regulations.

What this means for you

Pair each book with a current Academy guide (security, law, or agents) before operational decisions.

Selection Bias in Case Studies

Success stories hide failed deployments, reverted pilots, and hidden labor costs.

What this means for you

Ask “what would falsify this claim?” and run a small experiment before scaling.

Over-Indexing on One Author

A single worldview can anchor team strategy without diversity of risk perspectives.

What this means for you

Read contrasting book-club titles (optimist vs. critic) and document disagreements explicitly.