0.7.13 Human Judgment Is the Final Step

Tools Can Provide Information; People Make Responsible Decisions

Throughout FIT, you have used tools that provide evidence.

Examples include:

• compiler messages;
• debugger values;
• Feedback System results;
• spreadsheet summaries;
• charts;
• web responses;
• AI-generated output.

None of those tools removes the need for human judgment.

A person still needs to ask:

What does this evidence mean for the task?

AI Is One Evidence-Producing Tool

An AI system can generate:

• an explanation;
• a draft;
• a recommendation;
• a comparison;
• a list of possibilities.

The output becomes something you can examine.

It is not automatically the final decision.

A useful relationship is:

AI output
    ↓
human evaluation
    ↓
evidence and context
    ↓
human decision

Human Judgment Uses More Than the AI Response

A responsible decision can require information the AI does not have.

For example:

• current course requirements;
• workplace policies;
• local technical configuration;
• user needs;
• privacy restrictions;
• consequences of an error.

The person responsible for the task needs to bring those constraints into the decision.

Good Judgment Includes Knowing When You Do Not Know

Responsible human judgment is not:

I am a person, so my first opinion is automatically correct.

People can also:

• miss information;
• make assumptions;
• misunderstand evidence;
• rely on outdated knowledge.

A strong human decision includes the ability to say:

I need better evidence before deciding.

That is why Verify is part of the decision framework.

Expertise Changes What You Can Evaluate

A beginner may recognize that a generated C# example looks plausible without being able to determine whether it is safe or appropriate for a complex production system.

That limitation matters.

When the consequence is significant, seek appropriate expertise or authoritative evidence rather than pretending to evaluate beyond your current knowledge.

Human judgment includes knowing when to involve another qualified person.

The Final User Is Responsible for the Context of Use

Suppose AI produces an email draft.

The AI does not know every relationship, policy, or consequence surrounding the message.

The person sending it should decide:

• whether the information is accurate;
• whether the tone fits;
• whether private information should be removed;
• whether the message follows the relevant expectations.

The final use occurs in a real context.

That context is part of the human responsibility.

A Tool Can Be Correct and Still Be Inappropriate

Imagine an AI accurately explains a complicated technique that the current course has not taught.

The explanation may be technically correct.

Using it in an assessment could still be inappropriate if the task is designed to measure concepts within a defined learning boundary.

Correctness is one factor.

Appropriateness is another.

Human Judgment Should Be Explainable

A strong decision can be justified.

For example:

I revised the response because the factual explanation was useful, but the original included an unsupported claim about the current software version.

or:

I rejected the response because it required confidential information that the AI tool was not authorized to receive.

The explanation connects the decision to evidence and requirements.

Judgment Improves Through a Repeatable Process

A useful process is:

1. Understand the task.
2. Identify the important claims or choices.
3. Consider the source of the information.
4. Check for missing context or bias.
5. Verify important facts when needed.
6. Consider privacy and rules.
7. Choose Accept, Revise, Verify, or Reject.
8. Take responsibility for the final result.

The process matters because confidence alone is not evidence.

The Main Principle

AI can support human work.

It should not replace the responsibility to understand, evaluate, and justify the result.

The final step is a human decision informed by evidence.