Three Main Interaction Modes of AI
Automation
Augmentation
Agency
The 4D Principles
Delegation — Assign tasks to AI that it can handle. Before delegation, clarify what tasks need to be done and clearly understand AI’s capability boundaries.
Description — When assigning tasks to AI, describe the task content precisely and constrain the expected outcomes. Otherwise, you’re essentially drawing random cards—wasting resources (not only computational resources but also human cognitive resources for decision-making).
Discernment
- Humans must judge whether AI-generated results are usable. This judgment depends on how familiar the person is with the task type. For example, a novice programmer and an experienced programmer using the same AI model with the same prompt may produce entirely different outputs: the novice may accept anything that “appears functional,” whereas the seasoned programmer will reject unusable outputs and request revisions.
- So, how do you determine whether a result is good or bad? You must establish evaluation criteria in advance. For instance, when evaluating whether an explanation for a question is high-quality, the following criteria may apply:
- Sources must be official designated materials—not general internet sources.
- Language must be concise and clear.
- Explanations should follow a “general-to-specific” structure: first state what concept or skill the question tests overall, then explain why each answer option is correct or incorrect.
- etc.
- Without such standards, accepting outputs merely because they “appear usable” leads to inconsistency and non-reproducibility.
Diligence — Users must act ethically; avoid misusing AI or generating meaningless content.
Human vs. AI
Attributes humans possess—but AI struggles to replicate:
- Critical thinking
- Judgment
- Creativity
- Ethical oversight