What Is Action Selection?

Action selection is the cognitive process by which an individual decides what to do at any given moment, balancing desires, goals, and available options. It's the brain's way of navigating life's endless possibilities, steering behavior in a coherent direction. How does your mind orchestrate this intricate dance of decision-making? Join us as we unravel the mysteries of the human brain's executive choices.
Geisha A. Legazpi
Geisha A. Legazpi

Action selection is a process involving how a designed intelligent system will react next to a given problem. It is usually a field studied in psychology, robotics, and artificial intelligence. Action selection is synonymous to decision making and behavioral choice. The data gathered is researched and broken down in order to be able to adapt it to artificial systems like robotics, video games, and artificial intelligence programming.

Much of the data in the life sciences can be observed and experimented upon to evoke a variable response. All living creatures have their own instinctual reaction to food, predators, and mates. Creating a controlled environment where the studied animals are observed to always perform different solutions to different problems provides researchers and programmers with a basis for the advancement of their study. This in turn has led researchers and programmers in trying to recreate those instinctual responses in a controlled manner.

Woman doing a handstand with a computer
Woman doing a handstand with a computer

For researchers and programmers, the most common questions used in action selection are focused on what to do thereafter and what happens next. The responses in turn can be recycled for a new batch of experimental action selection. Prime examples of action selection can be found in games and artificial intelligence programming. In computer games, it can be found in First-Person Shooters (FPS) like Halo and Counter-Strike. Creatures, a pet-based game, uses an artificial intelligence engine that can make its own decisions by adapting to tasks.

What makes action selection a unique field is that there is always a strict guide to follow to have an acceptable level of data. The guide would always be based on a subject that is patterned off of a human or animal. For most, if not all researchers and programmers, a subject will always need to be placed in a location where the environment is unpredictable and is always changing. The subject will also need to react on time while performing a number of tasks. It must also interact with real live human beings in order to bring in a randomizing factor.

With those many random factors and a strict guideline to follow, research is never ending since there will always be a different set of circumstances for each experiment. One primary factor that makes researchers and programmers study this field intensively is response time. With each successful experiment where the subject has learned an action, a different avenue of action will come up. This in turn makes for a more complex subject when compared to a previous version.

You might also Like

Discussion Comments


@pleonasm - I think the problem is simply that we don't know how to build a brain that has the capacity and the flexibility to really be intelligent. At the moment, action selection is always going to be limited to the point where a human can see through it because our wetware is better than any software we're capable of creating.


@croydon - I think the most sophisticated artificial intelligence design goes into the enemies and backup characters in first-person shooter games these days. Action selection isn't something that game makers really want to make all that obvious, because the goal is to get the characters to seem as independent as possible. If you try to design a game where the independence of the characters is at the center of attention all the time, it's much more difficult to keep up the illusion.

That's why the closest you really get to that kind of game is usually based around characters that are supposed to be closer to animal, rather than human, intelligence.


I really loved the game Creatures when I was a teenager and I remember that I really bought into the idea that the simulation was a kind of artificial life. You could only run one game at a time and if you started a new one, it would erase the old one. My sister decided she wanted to play my game and started a new one, which "killed" all my little creatures. I didn't talk to her for weeks, because I felt like she had actually murdered them.

I don't think that the artificial intelligence engine on that game was actually that sophisticated but it certainly had me fooled. I'd be interested to see what they can come up with these days.

Post your comments
Forgot password?
    • Woman doing a handstand with a computer
      Woman doing a handstand with a computer