AI FLUENCY · PHASE 2: APPRECIATION
Module 8 — Human Intelligence and AI | Lesson 031
How Can Humans Stay in Control When Working with AI?
Explore human agency, meaningful oversight, independent judgment, and responsible decision-making in an AI-enabled world.
Watch the visual lesson. Then explore the examples, frameworks, knowledge check, and reflection activity.
Watch Lesson 031 on YouTube →Introduction: Who Makes the Decision?
Artificial Intelligence can generate ideas, analyze information, compare alternatives, and recommend actions.
These capabilities can help people make more informed choices.
But they also raise an important question:
When AI suggests what to do, who should decide what actually happens?
In Lesson 029, we examined the differences between human intelligence and AI.
In Lesson 030, we explored how their different capabilities can complement one another.
Now we examine how humans can remain active participants and decision-makers when working with AI.
This brings us to the concept of human agency.
1. What Is Human Agency?
Human agency is the capacity to establish intentions, make choices, and take meaningful action toward goals.
It includes deciding what matters, considering available options, and acting with responsibility.
For example:
- A student chooses a field of study.
- A teacher decides how to support learners.
- An entrepreneur evaluates a business opportunity.
- A professional makes a decision affecting their work.
AI may provide information, suggestions, and assistance in each situation.
However, receiving AI assistance should not automatically mean surrendering personal judgment or decision-making authority.
Human agency means retaining meaningful influence over the goals, choices, and actions that shape our lives.
2. Five Dimensions of Human Agency
Human agency in AI collaboration can be understood through five connected dimensions.
🎯
1. Purpose
Decide which goals matter and why they are worth pursuing.
Ask: What am I trying to achieve?
🧭
2. Direction
Guide AI assistance with goals, context, expectations, and boundaries.
Ask: What help do I actually want?
🧠
3. Understanding
Question AI outputs, assumptions, uncertainties, and limitations.
Ask: Do I understand this recommendation?
⚖️
4. Judgment
Evaluate alternatives and accept, modify, or reject suggestions.
Ask: Is this appropriate for my situation?
🤝
5. Responsibility
Recognize accountability for decisions and consequences.
Ask: Who is responsible for the outcome?
Together, these dimensions help people remain active participants in human–AI collaboration.
They also remind us that effective use of AI involves more than receiving and accepting answers.
3. Everyday Example: Choosing a Career with AI Assistance
Imagine a college student trying to choose a future career path.
The student has several interests but is unsure which opportunities might fit their abilities and aspirations.
The student asks an AI assistant:
"Suggest possible career paths based on my interests, skills, and education. Explain the qualifications I should investigate."
AI generates several possible career directions.
It may also suggest skills to develop, questions to investigate, and general sources of career information.
What Does the Student Do Next?
Step 1 — Explore
The student considers the options suggested by AI.
Step 2 — Verify
The student checks qualification requirements, career information, and opportunities using reliable sources.
Step 3 — Consult
The student talks with teachers, career advisers, or relevant professionals.
Step 4 — Reflect
The student considers personal interests, values, circumstances, and long-term goals.
Step 5 — Decide
The student determines which options to pursue, modify, or reject.
AI supports career exploration. The student retains ownership of the career decision.
AI may be useful for generating possibilities, but it can misunderstand personal circumstances or provide outdated career information.
The student therefore remains an active decision-maker.
4. The Human Agency Framework
A simple framework can help us understand how AI assistance and human control can work together.
What do I want to achieve, and why?
↓
What information, ideas, or alternatives can AI provide?
↓
Are the suggestions reliable, appropriate, and aligned with my goals?
↓
What will I choose, change, reject, or investigate further?
HUMAN PURPOSE → AI ASSISTANCE → HUMAN EVALUATION → HUMAN DECISION
This framework emphasizes the continuing importance of human goals and judgment.
For complex systems, maintaining control may also require technical safeguards, organizational governance, and appropriate authority to intervene.
5. Human Control Must Be Meaningful
Does human oversight exist simply because a person approves an AI recommendation?
Not necessarily.
A person may click an approval button without understanding the recommendation or its consequences.
Meaningful oversight requires the conditions needed for a person to exercise real judgment.
Information
Enough relevant evidence and context to review the recommendation.
Competence
Sufficient knowledge or access to expertise to assess the output.
Time
A realistic opportunity to investigate and decide.
Authority
The ability to question, override, or stop the AI-supported action.
These conditions are especially important when AI contributes to high-stakes decisions involving education, employment, health, finance, or public services.
Oversight must be designed to match the risks and practical circumstances of the activity.
6. The Risk of Over-Reliance on AI
AI systems can produce fluent and convincing responses, even when those responses contain errors.
People may gradually become inclined to trust AI suggestions without sufficient evaluation.
One relevant risk is automation bias.
Automation bias is a tendency to favor recommendations from automated systems even when other evidence suggests they should be questioned.
A Simple Example
An AI assistant recommends a career path and describes it confidently.
The student accepts the recommendation without checking whether the required qualifications, work conditions, or opportunities match their circumstances.
The risk is not simply that AI may be wrong.
It is that the student may stop exercising independent judgment.
How Can People Reduce Over-Reliance?
- Ask what evidence supports an important recommendation.
- Check facts and requirements using reliable sources.
- Consider reasonable alternatives.
- Notice uncertainty and missing information.
- Seek qualified expertise when the consequences are significant.
- Remain willing to disagree with AI suggestions.
Human agency grows stronger when people use AI to support inquiry rather than substitute for their own thinking.
7. How Can AI Expand Human Agency?
AI assistance and human agency do not have to conflict.
AI may strengthen agency when it helps people:
- Access relevant information.
- Explore options they had not previously considered.
- Understand unfamiliar concepts.
- Clarify their goals and priorities.
- Compare potential consequences.
- Prepare more effectively for independent action.
For example, a person may use AI to understand the questions they should ask before making an important decision.
In that case, AI is not choosing the person's future.
It is potentially expanding their ability to make an informed choice.
The aim is not less human thinking. It is better-supported human thinking and action.
Key Takeaway
AI should expand human agency, not replace it.
Purpose · Direction · Understanding · Judgment · Responsibility
Human–AI collaboration is most meaningful when people retain the ability and authority to direct, evaluate, and make decisions.
Quick Knowledge Check
Try answering each question before opening its explanation.
What is human agency?
A. Always following AI recommendations
B. The capacity to establish goals, make choices, and take meaningful action
C. Allowing AI to determine every important decision
Reveal Answer
B. Human agency involves forming intentions, making choices, and acting meaningfully toward goals.
Which set correctly identifies the five dimensions of human agency explored in this lesson?
A. Speed, scale, storage, prediction, automation
B. Data, algorithms, models, prompts, outputs
C. Purpose, direction, understanding, judgment, responsibility
Reveal Answer
C. These five dimensions help people remain active participants in AI-supported activities.
What does automation bias mean?
A. Questioning an automated recommendation
B. Giving automated recommendations undue preference despite conflicting evidence
C. Comparing AI with another information source
Reveal Answer
B. Automation bias can cause people to accept automated suggestions when closer examination is necessary.
Which example demonstrates meaningful human oversight?
A. Approving every AI result automatically
B. Ignoring all AI assistance
C. Reviewing evidence, questioning suggestions, and retaining authority to intervene
Reveal Answer
C. Meaningful oversight requires the knowledge, time, and authority needed to evaluate and act on AI recommendations.
Think About It
Imagine AI recommends an important decision about your education, career, work, or everyday life.
How would you decide whether to accept, modify, or reject its recommendation?
- Choose one meaningful decision you are considering.
- Describe your goal and why it matters.
- Identify what assistance AI could provide.
- List the evidence you would need to verify AI suggestions.
- Explain how you would evaluate the alternatives.
- State who retains responsibility for the final decision.
Write a short paragraph explaining how you would use AI assistance while preserving your purpose, understanding, judgment, and responsibility.
What Have You Learned?
After completing Lesson 031, you should be able to:
- Explain human agency in the context of AI use.
- Identify five dimensions of human agency.
- Distinguish AI assistance from human decision-making.
- Apply the Human Agency Framework.
- Recognize the risk of automation bias.
- Explain why meaningful oversight requires information, competence, time, and authority.
- Describe how AI may support rather than replace human agency.
Module 8 — Human Intelligence and AI
Lesson 029: How Is Human Intelligence Different from Artificial Intelligence?
Lesson 030: How Can Human Intelligence and AI Complement Each Other?
Lesson 031: How Can Humans Stay in Control When Working with AI?
These three lessons develop a connected understanding of human–AI collaboration.
UNDERSTAND THE DIFFERENCES
↓
RECOGNIZE COMPLEMENTARY STRENGTHS
↓
PRESERVE HUMAN AGENCY
Coming Next: Lesson 032 completes Module 8 by exploring the conditions for meaningful human–AI collaboration.
Your AI Fluency Journey
Understand AI
COMPLETE
Recognize Possibilities
CURRENT PHASE
Use AI Effectively
Achieve Meaningful Outcomes
Phase 1 established the foundations of AI.
Phase 2 explores how AI can contribute to meaningful human possibilities.
Phase 3 will focus on practical AI use.
Phase 4 will focus on accomplishing real-world human goals.
Continue Your AI Fluency Journey
Previous Lesson: Human–AI Complementarity
Watch Lesson 030: How Can Human Intelligence and AI Complement Each Other? →
Complete Phase 1 — Awareness
New to AI Fluency? Explore six foundational modules and 24 lessons.
Complete Course Roadmap
Further Learning Resources
NIST — AI Risk Management Framework
Stanford Institute for Human-Centered Artificial Intelligence
AI Can Assist. Humans Must Remain Agents.
AI can provide information, ideas, and recommendations.
Human beings establish goals, evaluate alternatives, and exercise judgment.
The real opportunity is not surrendering human decisions to AI, but strengthening our ability to make meaningful choices.
