Skip to main content

Bioengineering | Bioscience | Computer Science | Statistics

Understanding the intelligence of decision making

New research into how humans and animals use complexity to make decisions offers some tantalizing insights into the nature of intelligence.

An international team of scientists including KAUST's Professor Jesper Tegner has used a numerical approach to explain the complexity of behavioral patterns of ants. © 2023 KAUST; Ivan Gromicho.

An objective tool that characterizes the intrinsic complexity of behavioral patterns shows that humans and fruit flies alike are able to synthesize complexity in the environment to make decisions, which goes beyond merely making predictions from observations.

This leap could offer a step closer to understanding intelligence, offering insight into what artificial intelligence systems lack.

“For an animal to exploit the environmental deviation from some equilibrium, it must go beyond probabilities, beyond merely calculating the frequency of moves and beyond statistical entanglement with the environment,” says KAUST computer scientist Jesper Tegnér.

“Animals clearly distinguish between environments of varying complexity, reacting accordingly in real time. To understand this, and to construct intelligent systems, we need tools to evaluate and quantify ‘intelligent behaviors.’”

Learn more about the team’s research in this short video explainer.

Tegnér, with Hector Zenil from Cambridge University and James Marshall from Sheffield University, used a numerical approach to approximate the complexity of decision-making algorithms in animals as a tool to understand this intelligence behavior.[1]

The team used algorithmic information theory to measure the processes produced by deterministic choices, employing a metric called Kolmogorov complexity to quantify simplicity versus randomness, and so distinguish between correlation and causation. Essentially, this measures the complexity of a decision-making process from the length of its shortest possible description.

The researchers then combined this with a measure of the amount of computation required to make a decision to estimate the “sophistication“ of the decision-making algorithm.

“By applying these tools, we found that animals, including fruit flies and rats, appear to have some unknown mechanisms to perceive and cope with different degrees of complexity in their environments,” says Tegnér.

“Beyond just coping, they can harness and utilize the environment in their internal processes and in how they implement their decisions. This ability to assess and quantify environmental complexity and form and execute appropriate decisions is at the core of intelligent systems.”

The tools provide a numerical measure of “intelligence“ that validates previous observational studies. For example: ants take longer to communicate more complex instructions; fruit flies do not just fly randomly in the absence of stimuli but seem to seek out stimuli resulting in more complex behavior; while rats respond to increasingly complex environments by switching from deterministic to random behavior.

“The tools introduced here could help understand human decision systems, particularly how humans perceive and utilize randomness and complex behavior, and ultimately contribute to the design of cognitive strategies to improve artificial intelligence and mimic natural intelligence,” Tegnér says.

  1. Zenil, H. Marshall, J., Tegnér, J. Approximations of algorithmic and structural complexity validate cognitive-behavioral experimental results. Frontiers in Computational Neuroscience 16, 956074 (2023). | article.
You might also like