The empirical approach to cognition

has traditionally been dominated by

classical probability theory

However, these approaches has been

unable to explain some cognitive processes

For example, the skill of remembering

consecutive digits cannot be quantified

by classical statistical tools

such as Shannon entropy

because they cannot recognize

the algorithmic nature of a sequence

that the mind can easily recognize

and recall

What was needed was a much more

powerful, theoretical framework,

and the introduction of new tools

based upon the theory of algorithmic probability

That is, the ultimate theory

of induction and inference,

formally connecting complexity and probability

by way of universal computation.

For example, if a sequence of events

can be encoded by a short algorithm,

then a mind can understand and compress

that sequence even if it appears random

to traditional statistical tools.

This means, that statistical tools,

such as Shannon entropy applied to

inputs, outputs and internal representations

are not suitable to describe or explain

sophisticated mental processes.

However, the theory of algorithmic probability

applied to human and animal cognition

provides the theoretical means

to study these sophisticated processes

and even explains what was

traditionally seen as positive

or negative biases in terms of

choices that a mind takes.

For example, humans will tend to

find patterns in random data

that the traditional view will mostly

account for in the form of biased decisions

based on past experiences rather than

as a consequence of algorithmic process

of the mind trying to find sense and logic

even when there is none.

We have developed these tools

based on the principle that

a mind is better explained not by

statistical processes, biased by past experiences,

but as an algorithmic probability estimator

that has evolved to learn and exploit

the algorithmic nature of the world.

Paradoxically, this means the mind

may actually behave less often

as an irrational being,

biased by personal experiences,

and more like a general sophisticated,

predictable algorithm.

Indeed, early cognition research

found analogies with computation.

But soon found such analogy

overly simplistic.

Today, however, the bridge

between the mind and computers

we have made is more subtle,

and more sophisticated.

The bridge consists a model of the brain

that conceives the mind

as an idealized universal algorithm,

capable of generating models.

Our conceptual framework

unveils evidence supporting

these mechanisms of cognition.

We have shown that

landmark experiments on animals

can be confirmed by

simply quantifying the complexity

of behaviour time series from

different studies.

For example, ants maximize communication

efficiency as a function of sequence complexity.

Fruit flies do not behave in the simplistic

ways that were previous thought.

Rodents harness the power of randomness

when outsmarted by artificial intelligence.

And, humans are most creative

at generating randomness when

they reach 25.

These kinds of discoveries were,

however, impossible to make before

with traditional tools and classical measures,

and they foreshadow algorithmic cognition

as an exciting new field of research to tackle

some of the tough pressing issues,

related to mental disorders

and neurodegenerative diseases.