Dual system of reasoning
Human reasoning is the cognitive process by which individuals make sense of the world, draw conclusions, solve problems, and make decisions based on available information, experiences, and logical principles.
It has been explored by scholars and philosophers over centuries. The way humans understand reasoning has evolved as different thinkers introduced new perspectives, enriching the concept with insights from philosophy, psychology, and modern science.
If you want to know more, more in Knowing more section.
Today, reasoning is viewed as a complex cognitive process shaped by logical principles, sensory experience, cognitive biases, and evolving technologies.
Current reasoning paradigm agree upon dual system which we will drill little bit in order to help us understand it from our point of view.
Without further ado, lets make our simplified model for our quest. Or use one appropriate and just identify our tigers in it.
This is a principle/model how we gain knowledge, isn’t it?
When reading additional material from Knowing more section, I have found one scholar, Steven A Sloman, professor of Cognitive and Psychological Sciences at Brown University who study how people think: how we reason, make decisions, and form attitudes and beliefs.
In 1996 he published 20-page article in Psychological Bulletin, that integrates empirical findings from psychology and neuroscience in order to influence AI builders to make AI systems more cognitively realistic and capable of human-like reasoning.
It was a time the Turing test was still valid. 😊
One of the oldest conundrums in psychology is whether people are best conceived as parallel processors of information who operate along diffuse associative links or as analysts who operate by deliberate and sequential manipulation of internal representations.
Left or right system, system 1 or system 2, as we call them now?
An obvious solution to the conundrum is to conceive of the mind both ways — to argue that the mind has dual aspects, one of which conforms to the associationistic view and one of which conforms to the analytic, sequential view.
Such a dichotomy has its appeal: Associative thought feels like it arises from a different cognitive mechanism than does deliberate, analytical reasoning.
Sometimes conclusions simply appear at some level of awareness, as if the mind goes off, does some work, and then comes back with a result, and sometimes coming to a conclusion requires doing the work oneself, making an effort to construct a chain of reasoning.
These are exact his words (with all linguistic irregularities) and this sentences together with given table with categorization of basic characteristic of human reasoning (principles of operation, source of knowledge, etc.) is a perfect model of human reasoning we will use to better understand main concepts of cognition.
Can Machines Think — Take Two
In common life by reasoning we much more mean second (rule-based) system, because of causality and logic, kind of hard (irrefutable, unquestionable) conclusions we produce.
The output of the other side has no (logical) proof, just explanation, so it is often interpreted as gut feeling (not reasoning, not brain product).
Proposed dual model of reasoning enlarges our understanding of reasoning, simply by acknowledging that the gut feeling is a valid reasoning mechanism. With no proof (I insist!).
On the wider view, thinking is a very broad mental process that encompasses the generation and manipulation of ideas, concepts, and representations.
In this sense, reasoning is just a subset of thinking, thinking that has a reason, some specific outcome/decision it aims at.
And there are two ways of accomplishing it — associative and rule-based, intertwined and not equivalent at all (we will discuss those two systems in next column).
So, considering thinking and reasoning as well, we can easily conclude:
YES, machines can think (and reason through associative, not rule-based system).
Which was a huge surprise to me, because as former programmer, systems engineer, I have learned that machines (computers of that time) implemented only rule-based (if-then-else) systems of dealing with instructions, burned down into logical circuits from which all processors are made.
Pure logics, hardwired!
And yet the AI scientist and engineers managed to produce (almost) perfect associative system, characterized by the left part of our table. Of course, replacing personal experience with lot of (preprocessed) data in order to mimic kind of personal experience.
That is why we would not use it for some very specific and/or not publicly available information.
So far, we can easily agree that within our current mental model (constructed for our core purpose — to understand strengths and weaknesses of Gen AI and utilize them for our quest — personalized knowledge gathering) we have (almost) perfect companion which reasons by associative model, with no rule-based abilities.
They are (or should be) implemented in chatbot/application (communication) part.
If not, we have to take care of finding the reason why the given output is relevant, valid, land factual, which we would do with any other knowledge providers as well (books, articles, teachers, google…).
And if not, it is still valid knowledge, you yourself choose to believe into.
Future development
The whole Generic AI industry is fully aware of this lack of rule-based reasoning, implementing it through various methods.
As we speak, OpenAI launched OpenAI o1, formerly code-named Strawberry, the new release that have ability to reason through complex math, science, and coding tasks by evaluating its steps before proceeding — a major upgrade in explainability.
Knowing More
The Evolution of Human Thought on Reasoning
- Ancient Foundations: The Birth of Formal Reasoning
Socrates (469–399 BCE) initiated one of the earliest explorations of reasoning through his Socratic Method, which involved using dialogue and questioning to stimulate critical thinking and expose contradictions. Socratic reasoning was a form of dialectic — examining ideas through discussion to reach deeper truths.
Plato (427–347 BCE), a student of Socrates, added the notion of reasoning as a process of understanding the abstract, ideal forms. For Plato, reasoning transcended the material world; humans could only grasp true knowledge through intellectual insight, not sensory experience. This led to his idea that reasoning is a higher cognitive process beyond mere perception.
Aristotle (384–322 BCE), Plato’s student, brought significant advances by developing formal logic, which became the foundation of Western reasoning. He introduced the syllogism, a deductive reasoning structure where conclusions follow logically from two premises. Aristotle’s focus on logical reasoning aimed to bring structure and clarity to human thought, laying the groundwork for the systematic study of reasoning.
2. Medieval Era: Integrating Faith and Reason
During the medieval period, philosophers like Thomas Aquinas (1225–1274) combined Aristotle’s logical reasoning with Christian theology. Aquinas introduced the idea that reason and faith are complementary, showing that human reasoning could lead to an understanding of divine truths. This integration represented a key evolution, where reasoning was not just a secular tool but also a means to explore spiritual matters.
Aquinas contributed to the understanding that reasoning could exist within a framework of both natural knowledge (from logic and observation) and supernatural knowledge (from divine revelation), which was essential for medieval scholastic thought.
3. Enlightenment Era: Rationalism vs. Empiricism
The Enlightenment brought a new focus on reasoning as the primary means of acquiring knowledge, but there was significant debate about how reasoning should be understood.
René Descartes (1596–1650) was a central figure in the rationalist tradition. Descartes believed that reason, rather than experience, was the ultimate source of knowledge. He famously declared, “Cogito, ergo sum” (“I think, therefore I am”), establishing thought itself as the foundation of certainty. Descartes’ contribution was the emphasis on deductive reasoning and logic as the key to discovering truths.
In contrast, John Locke (1632–1704), representing empiricism, argued that all knowledge is derived from sensory experience. Locke rejected the rationalist idea that knowledge could exist independently of experience. Instead, he proposed that the mind is a tabula rasa (blank slate), shaped entirely by the information it receives from the outside world. Locke’s perspective added the idea that reasoning is not innate but developed through experience, leading to inductive reasoning.
Immanuel Kant (1724–1804) later attempted to bridge the gap between rationalism and empiricism. He argued that while knowledge begins with experience, human reasoning applies innate structures to organize this information. Kant introduced the idea of synthetic a priori knowledge, where the mind imposes structure on the world through categories of understanding, such as time and space. Kant’s contribution added a more nuanced view of reasoning, acknowledging both sensory experience and mental frameworks.
4. 19th Century: The Emergence of Psychological Perspectives
In the 19th century, reasoning began to be studied not just philosophically but also as a psychological phenomenon.
Wilhelm Wundt (1832–1920), the father of experimental psychology, shifted the study of reasoning from philosophical debate to scientific inquiry. Wundt viewed reasoning as a mental process that could be observed and measured, paving the way for empirical studies on how humans think and make decisions.
Sigmund Freud (1856–1939) introduced psychoanalysis, offering insights into how unconscious processes affect reasoning, particularly in decision-making and problem-solving.
Charles Darwin (1809–1882) introduced a biological perspective, suggesting that reasoning and human cognition evolved as adaptive functions. His theory of evolution influenced later psychologists who viewed reasoning as shaped by natural selection, as a means for survival and problem-solving in changing environments.
5. 20th Century: Cognitive Science and Artificial Intelligence
In the 20th century, reasoning became an interdisciplinary field, drawing from cognitive psychology, neuroscience, and computer science.
Jean Piaget (1896–1980) introduced a developmental perspective, showing how reasoning evolves in children. He proposed that reasoning develops in stages, from concrete operations to formal operations, where abstract thinking becomes possible. Piaget’s work highlighted that reasoning is not static but develops through interaction with the environment.
Herbert Simon (1916–2001) and Allen Newell (1927–1992) further revolutionized the study of reasoning by applying computational models to simulate human thought processes. Their work in artificial intelligence (AI) emphasized that reasoning could be broken down into algorithms — specific rules that computers can follow. Simon’s concept of bounded rationality argued that human reasoning is often limited by cognitive constraints, and we rely on heuristics (mental shortcuts) to make decisions in complex, uncertain environments.
Daniel Kahneman (1934–2024) and Amos Tversky (1937–1996) added an important dimension by exploring how reasoning is influenced by biases and heuristics. Their research into prospect theory showed that people’s decisions are often irrational, shaped by cognitive biases rather than pure logic. This added complexity to the understanding of human reasoning, emphasizing that it is not always optimal or rational.
6. Contemporary Approaches: Neuroscience and AI
Modern approaches to reasoning are shaped by advancements in both neuroscience and artificial intelligence. Neuroscientists now use imaging technologies to observe the brain’s activity during reasoning, revealing that it is deeply influenced by emotions and social contexts, not just logic. Dual-process theory, developed by researchers like Keith Stanovich and Jonathan Evans, posits that reasoning occurs in two systems: System 1, which is fast and intuitive, and System 2, which is slow and analytical.
In parallel, AI research continues to explore reasoning through machine learning and deep learning, where algorithms attempt to mimic human-like decision-making processes.
However, machine reasoning still lacks human intuition and ethical judgment, underscoring the complexity of replicating human reasoning fully in machines.
In Search for Knowledge publication
Mastering Insightful Dialogue with Gen AI
<PREV Invert, always invert
NEXT> Weapons of Mass Deception
