DEV Community

shangkyu shin
shangkyu shin

Posted on • Originally published at zeromathai.com

Concept of Artificial Intelligence: Rational Decision Making and Expected Utility Explained

Artificial Intelligence is often explained as machines that “think like humans.”

That’s not wrong—but it’s not the full story.

Cross-posted from Zeromath. Original article: https://zeromathai.com/en/concept-of-ai-en/


The Real Core of AI

Modern AI is not about imitation.

👉 It’s about making the best decision under uncertainty.


The Key Idea: Expected Utility

AI systems evaluate actions like this:

Expected Utility = probability × outcome

Then:

👉 choose the action with the highest value


Simple Example (Umbrella Problem)

  • Rain probability: 60%
  • No rain: 40%

Choices:

Take umbrella → +8

No umbrella →

  • rain: −20
  • no rain: +10

Expected Utility (no umbrella):

0.6 × (−20) + 0.4 × 10 = −8


👉 Result:

  • umbrella = +8
  • no umbrella = −8

👉 Rational decision = take umbrella


This Is Exactly How AI Works

Every AI system follows this loop:

  1. observe
  2. predict probabilities
  3. evaluate outcomes
  4. choose best action

Intelligent Agents

In AI, systems are modeled as agents:

  • perceive environment
  • take actions
  • maximize outcomes

👉 This is the core abstraction behind most AI systems.


AI vs Humans

Humans:

  • emotional
  • inconsistent
  • biased

AI:

  • probabilistic
  • optimized
  • consistent

👉 AI can be more rational than humans.


Where This Shows Up

This framework appears everywhere:

  • recommendation systems
  • reinforcement learning
  • autonomous driving
  • LLM decision policies

Final Takeaway

AI is not about copying humans.

It is about:

👉 optimal decision making under uncertainty


Discussion

Do you think current AI systems are truly “rational”?

Or are they just approximations?

Curious to hear your thoughts 👇

Top comments (0)