The study of human reasoning and argumentation has undergone a profound transformation over the past two decades, that being a shift from a normative, “logicist” paradigm (traditional theoretical framework that defines "good" reasoning by its adherence to the formal rules of logic, probability, and decision theory) to a descriptive, interactionist framework. Historically, the psychology of reasoning was dominated by an "intellectualist" view, which posited that the primary evolutionary function of human reason was to enhance individual cognition, that is, to allow the solitary thinker to navigate the world, correct false intuitions, and arrive at objective truths.¹ Under this traditional model, the failure of human subjects to adhere to the canons of deductive logic—such as the surprising widespread inability to apply modus tollens (or recognize it as valid) or the pervasive influence of confirmation bias—was viewed as a defect, a "flaw" in the cognitive system, or a failure of education.¹

However, a converging body of research from cognitive psychology, evolutionary biology, and developmental science has catalyzed a "social turn" in the understanding of higher cognition. The emerging consensus, crystallized most prominently in the Argumentative Theory of Reasoning (ATR), suggests that reasoning did not evolve to help the lone individual process information, but rather to facilitate communication, cooperation, and persuasion within complex social groups.¹ On such a view, argumentation is not a peripheral byproduct of general intelligence but the central adaptive function of the reasoning faculty itself. The aforementioned "flaws" are reinterpreted as functional specializations designed for the social division of cognitive labor.⁴

In this review, we'll bring together the most recent evidence in the field. We'll start by exploring the evolutionary and cognitive roots of argumentation, comparing the intellectualist view—with its focus on truth-seeking—with the interactionist perspective that maintains social and persuasive goals. Next, we'll look at how argumentation skills develop over time, particularly the ability to engage in genuine dialogue. After that, we'll address what actually motivates the content of our arguments: the role of motivated reasoning, why cognitive biases persist even among highly intelligent people, and how moral foundations shape what we choose to defend. Finally, we'll examine the psychology of online arguments, given ideas like the "backfire effect" and "echo chambers" in light of recent studies that have struggled to replicate earlier findings and consider potential solutions.

The central theoretical debate in the modern psychology of argument concerns the functional origins of reasoning. Why do humans argue, and why do they argue the way they do?

Proposed by Mercier and Sperber, the ATR gives a parsimonious evolutionary explanation for the structural features of human reasoning. The theory posits that reasoning is a social competence evolved for two distinct but complementary processes: the production of arguments to persuade others, and the evaluation of arguments produced by others.¹

A fundamental prediction of the ATR is a fundamental asymmetry in cognitive effort and bias, often summarized as "lazy production" and "vigilant evaluation."

This asymmetry solves the "Enigma of Reason".² If reason were designed for solitary truth-seeking and solitary truth-seeking alone, confirmation bias would be a catastrophic design flaw. However, if reason is designed for social argumentation, confirmation bias (or "myside bias") becomes a feature insofar as it ensures that each participant in a debate vigorously advocates for their perspective in a way that maximizes the group's exploration of the problem space.⁴

The interactionist hypothesis is strongly supported by empirical data on group problem-solving. In classic studies utilizing the Wason Selection Task, which tests deductive reasoning by making participants identify which cards to flip to test a rule. Individuals working alone typically perform dismally, with success rates often below 10%.³ However, when Moshman and Geil assigned the same task to groups, success rates skyrocketed to approximately 75%.³

Crucially, this improvement was not a result of the "smartest person in the room" dictating the answer or something to that effect. Analysis of group transcripts revealed that even when no individual member started with the correct solution, the group was able to construct the correct answer through the friction of debate. One member might propose a partial insight, another would critique it, and a third would synthesize the views. The interaction facilitated a level of performance that exceeded the cognitive capacity of the individual components.³

Of course, despite its explanatory power, the ATR faces critiques of its own. Developmental psychologists, such as David Moshman, argue that the theory may oversimplify the developmental trajectory of reasoning. Moshman suggests that while the ATR explains the origins of reasoning, it struggles to account for the sophisticated, metacognitive development of argumentation skills observed in late adolescence and adulthood, where individuals learn to internalize the critical process and reason logically in isolation.⁷

Furthermore, philosophers like Johnson have pointed out that the theory must account for the high-level reasoning of scientists and philosophers who work in relative isolation. Mercier and Sperber respond that such "solitary" reasoning is actually a simulated dialogue; the reasoner constructs an internal interlocutor, anticipating objections and refining arguments as if they were in a public debate.⁸ This internalization process is distinct from the innate capacity for social argument and represents a culturally scaffolded skill. See the table below for comparison between the two views

Alright so the cognitive architecture for rudimentary argument is innate, as it were, but the skill of high-level argumentation (the ability to distinguish evidence from explanation and to engage in dialogic counter-argumentation) follows a protracted developmental trajectory. The seminal work of Deanna Kuhn and colleagues has been instrumental in mapping this development.

Cross-sectional research indicates that argumentative competence does not mature automatically with age. In early adolescence (13-14 years), initial argumentative dialogues are characterized by "exposition" which is the articulation of one's own position without direct engagement with the opponent's claims.⁹

Kuhn observes that novice arguers often engage in a form of verbal parallel play. They state their thesis, wait for the opponent to state theirs, and then continue with their own line of reasoning, ignoring the content of the opposing turn.¹⁰ A persistent developmental hurdle is the conflation of explanation (saying why something happens) with evidence (providing data that proves it happens). Novice arguers frequently offer a causal narrative in place of empirical evidence, failing to recognize that a plausible story is not proof.¹¹

The development of argumentation is related to the maturation of our epistemological understanding, in other words, the individual's beliefs about the nature of knowledge.¹² The outlined progression through three stages looks like this:

Kuhn’s experimental work demonstrates that "dense exercise" of argumentative thinking is required to move students from exposition to true dialogue. In a series of longitudinal interventions, students who engaged in electronic dialogues with peers regarding social issues showed significant transfer of skills to individual writing tasks.^13. meaning those online arguments are worth something!

The intervention works because the social demand of a dialogue partner forces the individual to do more legwork than just generating claims (production) and makes them have to anticipate counterarguments (evaluation). Over time, this external social dynamic is internalized. The student learns to ask, "What would my opponent say to this?" even when writing alone.¹⁰ Now, I don’t want to overstate anything, but it would seem these findings suggest that the most effective way to improve critical thinking and writing is not through instruction in logic or essay structure, but through the practice of dialectic exchange. By arguing with others, we learn to think.¹⁵ This does not seem crazy at all to say in fact, it’s another reason why philosophy is so important as a discipline and why it should be implemented in more classrooms

A pervasive question in the psychology of argument is why individuals maintain beliefs in the face of contradictory evidence. Why is belief-preservation so strong? The literature offers two primary, competing frameworks: Motivated Reasoning and Cognitive Laziness.

Ziva Kunda’s seminal conceptualization of motivated reasoning proposes that motivation affects reasoning through a biased set of cognitive processes. When individuals possess a "directional motivation" (a desire to arrive at a specific conclusion), they selectively access memory and cognitive rules that support that desired outcome.¹⁷ Kunda emphasizes that motivated reasoning is not hallucination in the sense that they can come to any conclusion they want by pure volition. People can only believe what they can reasonably justify to themselves. They search for "seemingly reasonable justifications" to maintain their desired beliefs.¹⁷ The "illusion of objectivity" is maintained by the biased selection of evidence and that allows the individual to feel rational while being driven by desire.¹⁹ This bias is particularly potent when the belief is tied to identity or group affiliation. The motivation to maintain consistency between one's identity (e.g., "I am a smart, moral person") and one's past choices drives the rejection of threatening arguments.¹⁹

A critical finding and deeply interesting finding, by Keith Stanovich, is that confirmation bias, the tendency to evaluate evidence in a manner biased toward one's own prior beliefs, is largely independent of general cognitive ability (intelligence/IQ).²⁰ Indeed, one might assume that smarter people are less biased; yet, Stanovich’s data reveals that while intelligence correlates with the ability to generate complex arguments, it does not correlate with the ability to decouple from one's own perspective. In fact, high-IQ individuals may use their cognitive power to construct more elaborate rationalizations for their biased beliefs.²² This dissociation has led to the proposal of a "Rationality Quotient" distinct from IQ. Rationality involves epistemic regulation and the disposition to engage in "actively open-minded thinking" (AOT), whereas IQ largely measures algorithmic processing speed and capacity.²⁴

Recent research by Pennycook and Rand posits a theory contrary to the dominance of the motivated reasoning account, particularly in the context of "fake news" and political misinformation. They argue that susceptibility to false arguments is driven more by cognitive laziness (a lack of analytic thinking) than by partisan motivated reasoning.²⁵ Their studies utilize the Cognitive Reflection Test (CRT) to measure analytic tendencies. Results consistently show that individuals who score higher on the CRT are better at discerning real from fake news, regardless of whether the news aligns with their political ideology.²⁵ This suggests that people do not necessarily need to be "de-biased" (a difficult motivational intervention) but rather "nudged" to engage their analytic faculties. When people are encouraged to stop and think, their accuracy improves, even if the truth contradicts their political "side".²⁸

It is also the case that argumentation is often laden with moral weight. Moral Foundations Theory and Dyadic Morality seek to understand this

Haidt and colleagues developed MFT to explain the cross-cultural and political variance in moral reasoning. They identify at least five innate psychological systems: Care/Harm, Fairness/Cheating, Loyalty/Betrayal, Authority/Subversion, and Purity/Degradation.²⁹ A robust finding is the asymmetry in foundation usage. Liberals primarily rely on the "individualizing" foundations (Care, Fairness), while conservatives rely on all five, including the "binding" foundations (Loyalty, Authority, Purity).³⁰ According to them, this creates a "moral language barrier." An argument framed solely in terms of harm or rights (e.g., "Drug use harms no one") may fail to register with an interlocutor who processes the issue through the lens of Purity or Authority. And so effective persuasion requires "reframing" arguments to appeal to the target's moral foundations.³²

Gray and Wegner give their own alternative to MFT: The Theory of Dyadic Morality. They argue that the cognitive template for all moral judgment is fundamentally dyadic, requiring a perceived Moral Agent (intentional actor) and a Moral Patient (vulnerable victim).³⁴ TDM posits that all moral judgments are ultimately transformations of harm. "Purity" violations are often rationalized by inventing a victim (e.g., "it hurts society," "it hurts the soul"). People have an automatic drive to complete the dyad.³⁶ If we see a "victim" (Patient), we automatically search for a perpetrator (Agent). Conversely, if we see something "evil" (Agent), we search for a victim. This "dyadic loop" fuels conspiracy theories and polarization, as people act to protect perceived victims from perceived villains.³⁷ In argumentation, this manifests as a refusal to accept nuance; admitting the "villain" has a point feels like betraying the "victim."

Skitka’s research distinguishes between strong attitudes and moral convictions. When an attitude becomes a moral conviction, it is perceived as an objective, universal truth rather than a personal preference.³⁸ Moral convictions operate as "firewalls" against compromise. Research shows that while people usually value "procedural justice" (fairness of the legal/argumentative process), this preference evaporates when moral convictions are at stake. Individuals with high moral conviction care only about the outcome, not the process.⁴⁰ People perceive their moral convictions as derived from facts and reason, even when they are rooted in emotion. This "introspection illusion" makes them impenetrable to counter-arguments, which are dismissed as not only wrong but evil.⁴¹

Social media, the place we love and of course the melting pot of our debates. Which typically is a never ending hell, it would seem. Anyway, the transition of argumentation from face-to-face interaction to online platforms has generated significant concern regarding "echo chambers," "filter bubbles," and the "backfire effect." However, a rigorous review of recent empirical evidence suggests that many of these concerns are overstated, while other, more subtle mechanisms drive online toxicity.

The popular narrative is that algorithms trap users in ideological bubbles where they never encounter opposing views. Data from Guess, Nyhan, and Reifler suggests otherwise. As it happens, most users have relatively centrist and diverse media diets. The "echo chamber" effect is definitely real nonetheless, but it is statistically limited to a small, highly partisan minority (approximately 5-10% of users).⁴³ Ironically, research indicates that offline social networks (neighborhoods, churches, families) are often more homogenously segregated than online networks. The internet, by virtue of its wonderful chaos, often exposes users to more cross-cutting viewpoints than their physical lives do.⁴⁵ That said, algorithms optimize for engagement and so user self-selection is the primary driver of content consumption. People actively choose to read what they agree with; the algorithm merely assists this preexisting psychological preference.

(See here for tangential reading to this: https://theforensicfunnel.com/p/people-are-persuaded-by-rational-arguments)

One of the most famous phenomena in the psychology of argument was the "backfire effect" (Nyhan & Reifler, 2010), where correcting a misconception allegedly increased belief in it.⁴⁸ Massive replication efforts (Wood & Porter, 2019) with samples of over 10,000 participants have failed to find consistent evidence for the backfire effect. Generally, when people are presented with factual corrections, they do update their beliefs toward accuracy.⁵⁰ The thing is though, the distinction between factual belief and behavioral/attitudinal change. An individual might accept the fact (e.g., "The candidate did lie about X") but still support the candidate due to tribal loyalty. This suggests that while factual argumentation is not futile, it is often insufficient to change political allegiances.⁵²

And so, you ask, “If echo chambers aren't the primary culprit, what drives online polarization?” Your answer (according to the literature) is Social Identity Theory (SIT) which stipulates that affective polarization, the phenomenon where partisans dislike the "other side" because of their group membership (as compared to their policy positions).⁵³ It can be said that online argumentation often functions as identity signaling rather than persuasion. "Trolling" and polarized language are used to gain status within the in-group rather than to engage the out-group.⁵⁵ More importantly, and what was most interesting is that recent studies on AI-generated content suggest that the "technological realism" of AI pushes users toward the peripheral route of persuasion such that they focus on surface cues rather than the central route (logic), potentially exacerbating susceptibility to misinformation.⁵⁷

Possibly, the most important section of this article, given the interactionist nature of reason and the persistence of bias, how can argumentation be improved? The field is moving toward structural and dispositional interventions.

Intellectual humility, to no surprise, is the virtue that will help us out the most here and is becoming a key variable (if not the most important variable) in constructive discourse. Seriously, a nonthreatening awareness of intellectual fallibility will change nearly everything.⁵⁸ High IH is associated with greater open-mindedness, less dogmatism, and a superior ability to evaluate the strength of evidence.⁵⁹ In leadership contexts, displaying intellectual humility fosters "thriving" and better performance in employees by modeling openness to new ideas and reducing the social cost of error.⁶¹

Also, interestingly, new approaches in Computer-Supported Cooperative Work (CSCW) focus on designing systems that encourage "bridging" rather than polarization. Researchers like Ovadya propose ranking algorithms that prioritize content that appeals to diverse groups (e.g., content liked by both Democrats and Republicans) rather than content that maximizes engagement within a single cluster. This "bridging" metric promotes mutual understanding.⁶³ Studies show that even controversial topics can be discussed constructively if the conversation is framed around relationship-building rather than "winning." Users often want difficult conversations but lack the digital affordances to have them safely.⁶⁵ (Indeed, yet another reaon to think of bloodsports debate and eristic discourse as something to be avoided)

Finally, the field is correcting its weird, WEIRD (Western, Educated, Industrialized, Rich, Democratic) bias by examining cross-cultural differences in argumentation. For example, consider that Western traditions valorize high-arousal, confrontational argumentation ⁶⁷, where, much on the contrary, Eastern traditions often prioritize "inductive biases" that value consensus and source reliability over logical validity.⁶⁸ Of course, this is not to motivate any reason to believe that Easterners cannot argue, in fact, a recent historical analysis challenges the idea that Easterners "cannot" argue. Texts like the Han Fei Zi reveal sophisticated syllogistic reasoning when contextually appropriate, suggesting that cultural differences lie in the norms of argument usage rather than the capacity for reasoning.⁶⁹

The psychology of argument has matured from a study of logical errors to a study of social functionalism. As it should, we have long known that arguments are an interpersonal endeavor as much as they are (or even more so) a logical endeavor. The accumulated evidence would conclude that human reason is an instrument of social persuasion. The "flaws" of bias and lazy production are, in fact, features that facilitate the efficient division of cognitive labor in a group setting.

A beautiful example, tangential to this, that is said perfectly and brilliantly by Gerd Gigerenzer:

Now of course, this evolutionary adaptation faces a mismatch in the modern environment. The mechanisms evolved for small-group, face-to-face dispute where accountability is high, and feedback is immediate and so they are often hijacked by the scale and anonymity of digital platforms. We are vigilant filters of argumentation in person, but we rely on peripheral cues online that can be easily manipulated.

The future of the field lies in the so-called "psychological transactions" of argument.⁷⁰ We must understand not just the logic of the claim, but the moral and identity-based language that is traded when we argue. If the natural tendency of groups is to polarize, but the natural function of argument is to converge on truth, then the solution, as it has always been, is to design the right environments that align our social instincts with our epistemic goals. Our disinterested rational enterprises are all we have, to reason is to truth-seek, and to do so effectively we must create the conditions to do so. If not, how else will we convince all of these stubborn brilliant minds?