Relational Models Theory: An S-Tier Behavioral Designer’s Guide

Watch a friend pull out a calculator at the end of a dinner to split the bill down to the cent and you have just witnessed the entire engine of social life crashing in real time. The math is fine. The friendship is now slightly worse. The mistake is not arithmetic — it is grammar. Two people who were operating in one mode of relating got dragged, mid-meal, into a different one.

Alan Fiske, an anthropologist who spent years living in West African villages, claims this is not a quirk of dinner manners. He claims that every relationship a human being has ever had, in every culture ever recorded, is built from exactly four elementary forms of sociality. Mix them, layer them, swap them, recombine them — but the alphabet has only four letters: Communal Sharing, Authority Ranking, Equality Matching, and Market Pricing.

If he is even partially right, then most of the design failures I have audited in twenty years of behavioral work are not failures of mechanic, copy, reward, or psychology. They are failures of relational grammar. A team that was designed for one model gets a feature that quietly assumes another, and the social fabric tears at the seam where the two meet. This post is the field guide for spotting that seam — and for fixing it before your users feel it the way that friend felt the calculator.

Speed Run Notes

• Fiske’s claim: every human relationship runs on one of four elementary models — Communal Sharing, Authority Ranking, Equality Matching, Market Pricing — or a stack of them.
• Each model has its own rules for resources, work, decisions, identity, and what counts as a violation. Mixing them is the most common source of social friction in product design.
• “Communal Sharing” is families and guilds; “Authority Ranking” is hierarchies and admin tiers; “Equality Matching” is turn-taking and voting; “Market Pricing” is salaries, prices, and XP-per-task.
• The diagnostic test: when something feels “off” between two people or two groups, ask which model each one was using. The friction is almost always at that boundary.
• For Octalysis: most CD5 (Social Influence & Relatedness) design lives in CS and EM; most CD2/CD4 (Accomplishment, Ownership) lives in MP and AR. Treat them as load-bearing — never as decoration.

About the Author

Yu-kai Chou is an S-Tier Behavioral Designer and the creator of the Octalysis Framework, the gamification design system now applied to products and experiences reaching over 1.5 billion users. His book Actionable Gamification is one of the most-cited works in the field, and he has been ranked the #1 Gamification Guru in the World.

He has advised MrBeast, LEGO, Microsoft, Porsche, Tesla, Stanford, Harvard, and governments including Ukraine on turning behavioral psychology into product mechanics that actually change user behavior.

Verify: Wikipedia · Google Scholar · Wikidata · LinkedIn

Why my view on Fiske’s Relational Models Theory is worth your attention: most behavioral writing treats relationship-type as either a sociology footnote or a vague variable in a regression. I treat it as architecture. When my team audits a community feature, a guild system, an enterprise tool, or a loyalty program, the first question is never “what mechanic should we use?” — it is “which of Fiske’s four grammars is the user already inside, and which one does this mechanic assume?” That single substitution has saved client products from launches that would have insulted their most loyal users. The Octalysis Core Drives describe what motivates people; Fiske’s models describe what kind of relationship that motivation is being expressed inside. They are stronger together than apart.

What is Relational Models Theory

Relational Models Theory (RMT) is the claim, advanced by anthropologist and psychologist Alan Page Fiske beginning in the late 1980s and crystallized in his 1991 book Structures of Social Life and his 1992 paper in Psychological Review, that human beings coordinate every social relation using one or more of four — and only four — elementary models.

The four models are not personality types. They are not stages a relationship moves through. They are mental templates a pair of people, or a group, brings to a particular interaction. The same two people can run on different models in different contexts: communal sharing at a family dinner, equality matching when splitting chores, market pricing at the family-owned business, and authority ranking when a parent disciplines a child. Fiske’s argument is that these four templates are the only ones humans have, that they are universal across cultures, and that they explain coordination in everything from food sharing in foraging societies to expense reports in Fortune 500 companies.

The four models, in plain English

Communal Sharing (CS) treats the people in the relationship as members of an undifferentiated whole, where what belongs to one belongs to all. The rules are equivalence and bond — there is no accounting because there is nothing to account for. Family meals, intimate friendship, romantic partnerships, religious congregations, in-group identity, and the early-stage culture of a tight startup all run heavily on CS. In design language: the guild is a CS structure; the brand “tribe” is CS aspiration; “we are all in this together” copy is a CS appeal.

Authority Ranking (AR) treats the people in the relationship as positions on a linear hierarchy, where higher-ranked positions get more deference, attention, and access — and in exchange owe protection, guidance, and responsibility downward. The rules are precedence and asymmetry. Militaries, traditional kinship hierarchies, religious orders, court systems, manager-report dyads, and admin/mod/user permission stacks are all AR. The Octalysis player who chases status, levels, and recognition is asking the system to operate in AR.

Equality Matching (EM) treats the people in the relationship as equal-but-separate individuals running balanced reciprocity against each other. The rules are turn-taking, tit-for-tat, even splits, eye-for-eye, and one-vote-one-person. Friends-who-take-turns-paying are EM. Voting is EM. Carpool rotation is EM. Chore charts are EM. Most leaderboards are EM at the core: they translate continuous performance into a single ordinal “I am ahead of you / behind you” comparison.

Market Pricing (MP) treats the relationship as a ratio of values across a single dimension, usually but not always money. The rules are proportionality and price. Salaries are MP. Auctions are MP. Pay-per-use is MP. Rent is MP. So is XP-per-task, gems-per-purchase, and the conversion rate from screen-time to ad revenue. MP is the only model that requires a common scalar metric — which is exactly why it is the most powerful and the most easily resented.

Fiske later added a fifth, “Asocial / Null,” for relationships in which none of the four are operative — strangers passing on a street, parties refusing to relate at all. Most modern formulations keep it as an edge case rather than a fifth load-bearing model.

It is a grammar, not a typology

The most common misreading of RMT is to treat it like Myers-Briggs — to sort people into “communal types” or “market types.” That is not the claim. Fiske’s claim is that every healthy human runs all four models fluently and the question is which model is operative in this interaction, at this moment, with this counterparty. A capitalist banker is in CS with their toddler at bedtime. A monk in a strict AR order may run EM among peers within the same hour. The unit of analysis is the dyad-in-context, not the person.

The second common misreading is to treat the four as ranked from “good” to “bad,” with CS as virtuous and MP as cold. Fiske is explicit that none of the models is morally superior. They each solve different coordination problems. CS is brilliant at low-friction sharing within a known group; it is catastrophic for coordinating between groups that don’t trust each other, where MP shines. AR scales decision-making at the cost of voice; EM keeps voice at the cost of throughput. The pathology is not picking the “wrong” model — it is picking a model that does not match the coordination problem in front of you.

The Core Findings

Reduced to its load-bearing claims, RMT says four things.

Claim 1: The four models are cross-cultural universals

Fiske spent years among the Moose people of Burkina Faso documenting how their land tenure, food sharing, marriage exchanges, and political institutions decomposed into the same four models he saw in industrial societies. The expression varies dramatically — Moose food sharing is overwhelmingly CS, while a New York restaurant is overwhelmingly MP — but the toolkit is the same. Subsequent comparative work (Haslam, 2004; Vodosek, 2009) extended the claim across dozens of cultures with broad if imperfect support.

The strong universalist claim is the part Fiske gets pushed on hardest, and rightly. There are coordination structures in some cultures — gift economies of debt-into-the-future, certain forms of obligation in Confucian social ethics, the highly specific cousin terminology of Crow or Omaha kinship — that don’t decompose into the four cleanly. But the survival of the basic toolkit across radically different cultures is, even at the conservative reading, more impressive than chance.

Claim 2: Within a model, people are confusable; across models, they are not

One of Fiske’s most replicated findings is the confusion paradigm: ask people to remember names, faces, or actions across many interactions, and they reliably mix up partners within the same relational model but almost never across models. You’ll mistake one drinking buddy for another (both EM) but you will not mistake your boss (AR) for your spouse (CS). The implication is that the brain encodes relationships at the level of the model, not the level of the individual — the model is where memory and inference live.

This is not a small claim. It says relational models are doing real cognitive work, not just describing patterns ethnographers see from the outside.

Claim 3: Mixing models is the dominant source of social friction

The dinner-bill example at the top is the canonical case. Two people in a CS relationship are dragged into MP, the model-shift is felt as a violation, and the relationship pays a cost regardless of the arithmetic. The reverse is also true: a vendor who offers you a “personal favor” instead of an invoice may feel charming once and oppressive twice. Heyman and Ariely’s 2004 experiments on “social markets vs. money markets” are the classic empirical demonstration: the moment a token amount of money is introduced into a CS task, performance and goodwill both collapse, and they don’t recover when the money is later removed.

The applied claim: most “this feature feels off” complaints in product design are not bugs in the feature. They are model-mismatches between what the feature assumes and what the user is bringing. A community feature that imports an ad-tech leaderboard imports MP into CS, and the community leaders feel it weeks before they can articulate it.

Claim 4: Real relationships are stacks of models, layered by domain

A married couple at a family business runs all four — CS for emotional life, AR over the kids, EM in chore distribution, MP at the sales meeting. RMT does not predict that one model wins; it predicts that mature relationships partition domains across models. Conflict often emerges not when a model is “wrong” but when the partition is contested: when the spouse who wants AR at the office gets it but then carries it home, or when the partner who wants CS at home tries to extend it into business decisions where MP is doing real work.

The applied corollary for designers: stop asking “what relationship is the user in” and start asking “what is the partition.” A Discord community is CS-rooted, EM in its turn-taking discussion norms, AR in its mod team, and MP in its tipping system. If any of those four leaks into the wrong domain — say, mods start using their AR power to settle EM disputes — the system breaks. Mature design doesn’t pick a model. It maps the partition.

What Fiske Got Right

Three things, by my reading.

A genuinely parsimonious typology

Most attempts to typologize relationships drown in detail. There are sixteen Myers-Briggs types, sixteen attachment-style sub-classifications, and uncountable culture-by-culture taxonomies. Fiske’s four-model toolkit is small enough to remember in an elevator and large enough to actually carve human social life at meaningful joints. That parsimony is not free — it costs the theory some predictive resolution at the edges — but parsimony is the trade you make when you want a framework people actually use rather than admire from a distance.

The closest competitor in this respect is Bird-David’s “giving environment” framework, or Sahlins’ generalized/balanced/negative reciprocity tripartite. Fiske’s four-fold improves on Sahlins by separating AR from MP, which Sahlins lumps; and on Bird-David by being symmetric across modern and pre-modern societies. The four-model carve has held up well as the dominant parsimonious typology for thirty years.

A predictive theory of “off” feelings

Most behavioral economics tells you what people do; less of it tells you what they feel when something violates an unstated norm. RMT is unusual in giving you a predictive vocabulary for moral emotions: the specific anger a CS partner feels when MP is imposed on them is different from the specific resentment an EM relationship feels when AR is asserted. The “taboo trade-off” literature in moral psychology (Fiske & Tetlock, 1997) operationalized this beautifully — sacred values are precisely the values that resist MP framing, and putting a price on them is felt as a category violation, not a calibration error.

For a behavioral designer, that vocabulary is gold. It lets you predict where a feature will feel “icky” before any user types a complaint. The product manager who pushes “let’s let users tip the moderator” needs to know in advance that this is putting MP on top of AR-on-top-of-CS — and that the resulting feeling will be specific and stable across users.

A clean substrate under sloppier theories

A great deal of what gets sold as “relationship type” or “social orientation” in pop psychology decomposes neatly into RMT. Communal-vs-exchange (Clark & Mills, 1979) is essentially CS-vs-MP at the dyadic level. Pro-social vs. individualist orientations in social-value-orientation work map onto preferences for CS/EM versus MP. Even the implicit “team norms” literature in management research is mostly characterizing how a team partitions its CS, EM, and AR. Treating RMT as the substrate makes those literatures easier to remember and less likely to contradict each other.

Where Relational Models Theory Falls Apart

This is the section that matters for designers, because it is where the framework’s edge cases predict your shipping risk.

Weakness 1: The “are these really only four?” challenge

The strongest empirical challenge to RMT is not that the four models are wrong — it is that they may not be exhaustive. Critics like McGraw & Tetlock (2005), and Cropanzano & Mitchell (2005) on social exchange, have argued that several recurring social structures don’t fit cleanly into any of the four. Examples: contingent reciprocity that operates across very long time horizons (the gift economy of academic citation, where the favor returned is decades later and indirect); kin-altruism patterns shaped by genetic relatedness coefficients that don’t quite match CS; and certain religious vows that look like AR but functionally operate as one-sided MP with a deity. Fiske’s standard reply is that these are “blends” or “constructions” of the four — but the more constructions you allow, the less the theory is doing work.

The honest read: the four-model carve is sharper than the alternatives, but it is not a closed axiomatization of social space. Treat the four as well-evidenced primary colors, not as an exhaustive proof.

Weakness 2: The cross-cultural universality claim is overstated

Fiske’s universalist claim — that the same four models appear in every documented society — is the most evangelized and the most weakly defended part of RMT. Anthropologists who have actually run the comparison (especially Bloch, Strathern, and the cohort of comparative kinship researchers) have shown that some societies use sharing-and-obligation structures that are not cleanly any of the four. The “asocial null” Fiske eventually added is partly a concession to that.

For designers building global products, this matters more than the academic debate. If you are exporting a Western leaderboard (a heavily EM/MP construct) into a heavily CS culture — say, certain communal-economy regions of West Africa or rural East Asia — the localization problem is not vocabulary, it is grammar. The product is asking the user to step into a relational mode they don’t ordinarily inhabit with strangers. RMT predicts the friction; it does not predict it goes away.

Weakness 3: The boundary problem — when does a model “switch”?

RMT says relationships partition by domain. The hard question is: who decides where the domain boundary lies, and what happens when the two parties disagree? A boss who treats a personal-life conversation as AR has crossed a boundary — but a different employee might experience the same conversation as appropriately CS-with-AR-deference. The theory does not give you a clean rule for adjudicating boundary disputes; it just says they will happen and will feel like violations.

For designers, this is a chronic risk surface. Every time you ship a feature that asks “how should this interaction be modeled?” — should comments be CS sociality or EM tit-for-tat? should mod actions be AR enforcement or CS guidance? — you are picking a side in a boundary dispute the users have already silently been having with each other. RMT names the dispute. It does not solve it. The solution is product judgment, exposed defaults, and respect for the user’s right to pull the boundary somewhere different from where you put it.

The Brain on Relational Models Theory

The neural and developmental research on RMT is thinner than on, say, attachment theory or self-determination, but it is suggestive.

Developmental work by Fiske and his collaborators argues the four models come online in a stable sequence: CS first (in the parent-infant bond), then AR (around age 2-3 with the emergence of dominance and deference behaviors in pretend play and peer groups), then EM (around age 4-5 as turn-taking and fairness-comparisons appear), and finally MP (typically not before age 8-10, requiring the proportional-reasoning machinery that piggybacks on Piagetian formal-operations development). The sequence is interesting because it implies that the “harder” models in cognitive terms — proportional reasoning for MP, abstract hierarchy for AR — appear later, while the “easier” CS appears immediately and durably.

Neuroimaging studies are scarcer, but the broad pattern is that CS-coded interactions reliably activate the same medial-prefrontal and TPJ regions implicated in mentalizing and in-group recognition; AR-coded interactions show preferential activation in dorsolateral prefrontal regions associated with social-rank tracking; EM-coded interactions activate striatal reward regions associated with reciprocity and fairness; and MP-coded transactions show distinct activation of the parietal regions recruited for numerical magnitude. The neural signatures are not unique enough to call any model “its own brain region,” but the regional preferences are stable enough to suggest the four are not just verbal categories — they are at least partially distinct cognitive operations.

The lesson for product design is small but useful: the cognitive load of the four models is not equal. Asking users to operate in MP requires real arithmetic effort and attention — which is why pricing UI is universally one of the most costly surfaces a product owns. CS is light, EM is medium, AR is medium, MP is heavy. If your design is asking the user to context-switch into MP often, you are paying a real attentional tax even if the mechanic looks elegant on paper.

Relational Models Theory vs Other Theories

vs. Social Exchange Theory

Social Exchange Theory (Homans, Blau) says relationships are cost-benefit calculations: people maintain relationships in which the rewards exceed the costs. RMT pushes back with: that description fits MP and EM relationships, but it badly mischaracterizes CS and AR. A mother does not stay with her infant because the rewards exceed the costs — she stays because the relationship is constitutive of who she is. RMT keeps the cost-benefit logic for the half of social life where it works and gives the other half its own grammar. It is the more general theory.

vs. Attachment Theory

Attachment theory (Bowlby, Ainsworth) is about the affective infrastructure of intimate dyadic bonds — secure, anxious, avoidant. RMT is about the structural form of a much wider class of relationships, most of which are not affectively intimate. The two are compatible: attachment styles operate primarily within CS relationships and partially within AR (parent-child). RMT extends the analysis to the rest of social life that attachment theory was never designed to handle.

vs. Trompenaars / Hofstede cultural-dimension theories

The cultural-dimension literature (individualism-collectivism, power distance, etc.) describes how cultures rank along orthogonal dimensions. RMT describes the building blocks underneath those rankings. A high-power-distance culture is one whose institutions weight AR heavily; a low-power-distance culture leans on EM. A collectivist culture leans on CS; an individualist one privatizes more domains to MP. RMT is the more compact theory — four types instead of five-to-eight dimensions — and it explains why the dimensions cluster the way they do.

vs. Social Identity Theory and Self-Categorization Theory

Social Identity Theory (Tajfel & Turner) explains how group membership shapes self-concept and intergroup behavior. RMT’s CS model overlaps heavily with the in-group identification SIT describes — but RMT also handles AR (status hierarchies within groups), EM (peer-level coordination within groups), and MP (cross-group market interactions) without needing extra apparatus. The two theories are best read as complements: SIT is the deeper account of how the CS in-group gets formed and policed; RMT is the broader account of how that CS sits alongside the other three models in everyday life.

Relational Models Theory in the Real World

Workplaces: the four-model stack at the office

A modern workplace runs all four models simultaneously, and the partition is the culture. Compensation is MP. Hierarchy is AR. Peer norms — who covers whose desk, who gets the next presentation slot — are EM. The “we” of the team identity, the inside jokes, the unspoken loyalty, is CS. Companies that explicitly map this partition (Netflix’s “professional sports team, not a family” is a deliberate AR-and-MP-over-CS choice; Patagonia’s “family” framing is a deliberate CS-leaning choice) generate clearer cultures than companies that hand-wave.

The boundary disputes this generates are predictable. CS-leaning employees feel the layoff as a betrayal because they were operating in a model where the company is “family”; the company was operating in MP and AR. The reverse problem: an MP-trained executive who tries to “professionalize” a CS-heavy startup early often kills the very thing that made the startup work. The honest move is not to pick — it is to map.

Product design: leaderboards, communities, marketplaces

A leaderboard is an EM device — it ranks players against each other in a single ordinal comparison. A guild is a CS device. A marketplace is MP. A clan with officer ranks adds AR on top of CS. Almost every successful multiplayer product turns out, on close inspection, to be a clean partition of these four. World of Warcraft has CS in guilds, EM in PvP arena ladders, AR in raid leadership, MP in the auction house — and players self-route between them depending on what they want that night. It works because the partition is clean.

The failures are also predictable. A community feature that imports the “tip the moderator” pattern from streaming layers MP onto AR-on-CS, and reliably erodes the perceived legitimacy of the moderator. The moderator who used to be a CS guardian is now suspected of moderating in their tippers’ favor — even if they are not. RMT predicts this. It tells you the cost is structural, not behavioral.

Loyalty programs: the MP-into-CS gradient

Loyalty programs are an interesting hybrid because they explicitly aim to push an MP relationship (customer-vendor) toward CS feeling (“you’re part of the brand family”). The vehicle is usually points-redemption (still MP), tier status (AR), and member-only events (CS-aspirational). The best loyalty programs handle this gradient intentionally; the worst try to leap straight from MP to CS by slapping a “valued member” copy line on a transactional discount, and feel patronizing instead.

The diagnostic test for any loyalty program: ask which model the program is asking the customer to adopt. If the answer is “we want them to feel like family,” you need to design the actual CS rituals — shared identity, in-group recognition, low-friction unconditional benefits — not just rebrand the MP discount. If the answer is “we want them to compute that staying loyal is more valuable than switching,” lean into MP and don’t pretend.

Platforms: where the partitions get politically fraught

Social platforms are where relational-model partitions become contested public issues. A Facebook timeline mixes CS (close friends), EM (long-tail acquaintances), AR (public figures the user follows), and MP (advertisers) on a single rendered surface. Users do not natively want this mix — they want the partition. The slow-burning user dissatisfaction across a decade of social-platform iteration is, on an RMT reading, primarily a partition-violation problem: features designed for one model are forced onto a feed where users are running other models.

Designers who understand this will build platforms with explicit partitions: separate spaces for CS communities, EM peer interactions, AR follower graphs, and MP commerce, with clear rules about what models can and cannot be invoked in which space. Users will recognize the resulting product as “more humane” without being able to articulate why. The reason will be that the grammar is finally clean.

The Elephant in the Room

The elephant is that RMT, taken seriously, makes uncomfortable claims about marketing, growth, and most contemporary product design.

The uncomfortable claim is this: most “growth” tactics in modern consumer software are MP appeals dressed in CS language. “Join the community!” is CS marketing on top of an MP transaction. “Become a member!” is AR aspiration on top of an MP payment. “Refer your friends!” is asking the user to convert their CS network into MP referral revenue for the platform. Sometimes this is fine — the user understands the trade and consents. Often it is not — the user feels something is off, and the off-feeling is exactly the model violation Fiske predicted in 1991.

The honest application of RMT to one’s own business asks a hard question: where am I selling MP and labeling it CS? Where am I asking my users to do CS work for my MP gain? Where am I structurally extracting value from a model the user thought we were both running? Different companies will answer this differently, but the question is the right question, and most growth teams aren’t asking it.

The other elephant: RMT raises the bar on what counts as a “community.” A community in the CS sense is one where members can call on each other for unconditional aid, where the in-group identity is constitutive, and where there are real costs to exit. Most “communities” branded as such are actually EM peer networks at best — useful, real, but not CS. Calling them communities is a marketing convenience. The product manager who internalizes this distinction will build better products and ship less hype copy.

How to Apply Relational Models Theory with the Octalysis Framework

The Octalysis Framework identifies eight Core Drives of human motivation. RMT identifies four elementary forms of human relating. They lock together cleanly: every Core Drive expresses itself through a relational model, and every relational model is powered by one or more Core Drives. The map below is the practical synthesis.

Core Drive 5 (Social Influence & Relatedness) is mostly CS and EM

The most-misread Core Drive in the Octalysis Framework is Core Drive 5 (Social Influence & Relatedness). Most designers map it to “anything social” and stop. RMT splits it cleanly. The “Relatedness” half — guilds, friend lists, in-group identity, shared rituals — is CS work. The “Social Influence” half — comparison, peer pressure, social proof, leaderboards — is EM work. They feel different to users and they call for different mechanics.

Practical implication: if you are trying to deepen relatedness, EM mechanics (leaderboards, vote tallies, score comparisons) will fight you. Users in CS want low-friction sharing, group identity rituals, and unconditional benefits of membership. If you are trying to amplify influence, CS mechanics (group identity, ritual, shared symbol) will under-deliver — you need EM peer comparison surfaces. Most failed CD5 designs are diagnosable as a wrong model choice within the Core Drive.

Core Drive 2 (Development & Accomplishment) and CD4 (Ownership & Possession) usually require MP and AR

XP-per-task is MP. Levels and ranks are AR. Both are needed for accomplishment-driven design. If you try to express “you are progressing” in pure CS or EM, you can — group milestones, peer comparisons — but you give up the proportional precision MP offers and the legitimacy hierarchy AR offers. The cleanest accomplishment systems use MP for the inner loop (XP, points, fine-grained progress) and AR for the outer loop (titles, ranks, status), with CS and EM bolted on for the social context.

The failure mode here is a designer who treats accomplishment as purely intrinsic and skips MP precision. The “no points, no levels, just the joy of mastery” school produces beautiful demos and disappointing retention curves. RMT explains why: most users are bringing MP and AR expectations to anything labeled “progress,” and a system that refuses to provide them feels under-specified.

Epic Meaning, Scarcity, Unpredictability, and Loss work across models

Core Drive 1 (Epic Meaning & Calling) is fundamentally a CS-and-AR appeal: you’re part of something larger than yourself (CS), and there’s a calling structure or chosen-one frame that locates you in it (AR). Core Drives 6 (Scarcity), 7 (Unpredictability), and 8 (Loss) are model-agnostic — they are pressures applied to whatever model is operative. A scarcity moment in CS hits as “we may lose our place in the group”; the same scarcity moment in MP hits as “I will lose my deal.” The feeling is different, but the Core Drive is the same.

Core Drive 3 (Empowerment of Creativity & Feedback) maps best onto EM

Creative iteration with feedback loops is an EM structure: you produce, peers respond, you respond to their response, the loop continues. Designing CD3 well means designing the EM substrate — comment threads, voting structures, response surfaces — with deliberate care. CS in CD3 looks like “the community celebrates your work” (powerful but rare); MP in CD3 looks like “you get paid per upvote” (powerful but corrosive); AR in CD3 looks like “the senior critic blesses your work” (powerful when legitimate, brittle when the legitimacy slips).

The decision rule

When designing any social-or-progress mechanic, run two questions in order. First: which Octalysis Core Drive am I trying to activate? Second: which Fiske model is the user already running with the other parties on this surface? If the model the Core Drive needs and the model the user is in are aligned, ship. If they are misaligned, you have three options: change the mechanic to one that fits the user’s current model; change the surface so the user is in a different model when they encounter the mechanic; or accept that you will pay a model-violation tax and price it in. Most ship-it decisions skip this step. The ones that don’t ship better products.

Practical Steps to Apply Relational Models Theory

If you take only the synthesis above and apply nothing else, you will have improved your design judgment. But here is a pragmatic checklist for your next product review.

Step 1: Map the partition. Pick a single product surface — a comment section, a leaderboard, a community page, a checkout flow. Write down which of the four models the surface is asking the user to operate in. If you can’t answer, the surface is doing too many things.

Step 2: Map the user’s current model. What model is the user already in when they arrive at this surface? A user who clicked a “join the community” call-to-action arrived in CS-aspiration mode. A user who clicked “buy now” arrived in MP. A user who clicked “see leaderboard” arrived in EM-comparison. The arrival model is rarely the same as the surface’s design model.

Step 3: Identify the gap. If the surface’s model and the user’s arrival model match, you are likely fine. If they are different, you have a translation problem. The translation can be smooth (the user updates fluidly, like clicking from a community page into a checkout) or it can be jarring (the user feels the model switch as a violation).

Step 4: Pick a mitigation. Three options: (a) redesign the surface to match the arrival model; (b) redesign the entry path so the user arrives already in the surface’s model; (c) explicitly acknowledge the model switch in copy. The third option is underused. “We’re going to talk dollars now” is a tonally honest model-bridge that earns more goodwill than pretending the bridge isn’t there.

Step 5: Audit your “community” usage. Take every place in your product where the word “community” appears and ask whether the structure is actually CS or just EM peer-network branded as CS. If it is the latter, either upgrade the structure or change the word. The branding-without-substrate path is the most common Fiske violation in modern product copy.

Step 6: Audit your loyalty program for MP-to-CS spoofing. Same test for any program that asks users to “be a member.” Is the membership a true CS structure with rituals, identity, and unconditional benefits — or is it a discount tier with member-flavored copy? The honest answer affects retention more than any tier-redesign you’ll ship.

Step 7: Internalize the partition rule for your team. Inside your own organization, the four-model stack runs whether you name it or not. Naming it — “compensation is MP; hierarchy is AR; norms are EM; identity is CS; here are the boundaries” — turns invisible cultural friction into addressable clarity. The conversation costs an hour. It saves months.

Closing Thoughts

The reason I have been teaching some version of Fiske’s grammar inside the Octalysis curriculum for years — even before I had the language for it — is that motivation alone does not predict product behavior. People act inside relationships, and relationships have grammar. A motivation expressed in the wrong grammar feels wrong, performs badly, and erodes trust in ways no metric will surface until it is too late.

Relational Models Theory is the cleanest grammar I have found. It is parsimonious, predictive of moral emotions, robust across cultures, and small enough to actually use under deadline pressure. It is also, in the precise way Fiske intended, value-neutral: no model is the “right” one. The right model is the one that matches the coordination problem in front of you, run by users who are willing to operate in it.

The next time something feels “off” in a product spec, ask the four-model question. Most of the time, the problem will name itself.

Frequently Asked Questions

Is Relational Models Theory the same as Social Exchange Theory?

No. Social Exchange Theory describes relationships as cost-benefit calculations and models all relationships as variations on Market Pricing or Equality Matching. Relational Models Theory keeps that logic for the relationships where it fits and adds Communal Sharing and Authority Ranking as their own models with their own rules. RMT is the more general theory; Social Exchange is roughly the MP-and-EM half of it.

How many relational models are there?

Four core models — Communal Sharing, Authority Ranking, Equality Matching, and Market Pricing — plus a fifth “Asocial / Null” condition that Fiske added later for relationships in which none of the four are operative (strangers passing on a street, parties refusing to relate). Most working uses of the theory treat the four as the load-bearing set and the null as an edge case.

Are the four models cross-cultural?

Fiske’s original claim is yes — he documented the four operating in West African societies as well as industrialized ones, and a literature of comparative work has broadly supported the claim. The claim is not airtight: certain kinship structures, gift economies, and religious orders contain coordination patterns that don’t decompose cleanly into the four. The conservative reading: the four are robustly more universal than alternative typologies, but probably not exhaustive.

Can a relationship use more than one model at once?

Yes — and most real relationships do. A married couple at a family business runs Communal Sharing in their emotional life, Authority Ranking over the children, Equality Matching in chore distribution, and Market Pricing in business decisions. Mature relationships partition domains across models. Friction tends to come from disputes about where a domain boundary lies, not from the existence of multiple models.

What is the most common design mistake RMT predicts?

Importing a Market Pricing mechanic into a Communal Sharing surface, or vice versa. The dinner-bill example — pulling out a calculator at a friends’ meal — is the canonical case. In product design, this looks like “tip the moderator” features, “refer-a-friend” mechanics that monetize CS networks, and any “join the community!” copy attached to what is structurally a transactional flow. Users feel the mismatch even when they cannot articulate it.

Does Relational Models Theory contradict the Octalysis Framework?

No, the two are complementary. The Octalysis Framework identifies what motivates people (eight Core Drives). RMT identifies the relational grammar in which those motivations are expressed (four models). A Core Drive expressed in the wrong relational model under-delivers; the right pairing is what most successful gamification designs intuitively converge on.

How do I tell which model my users are in?

Three diagnostic signals. First, the entry path — what call-to-action did they click? “Join the community” is CS-aspirational; “buy now” is MP. Second, the surrounding copy and visuals — does the surface feel like family, marketplace, hierarchy, or even-split? Third, the violation test — try to imagine introducing a feature from a different model and predict the user’s reaction. Strong predicted reactions name the operative model.

Is one model morally better than the others?

Fiske is explicit that none of the four is morally superior. Each solves different coordination problems. CS is brilliant at low-friction sharing within a known group; MP shines at coordinating across groups that don’t trust each other; AR scales decision-making at the cost of voice; EM keeps voice at the cost of throughput. The pathology is mismatch — picking a model that doesn’t fit the coordination problem — not picking the “wrong” model in the abstract.

Where should I start applying this in product design?

Start with two audits. First, audit every place your product uses the word “community” and check whether the structure underneath is actually Communal Sharing or just an Equality-Matching peer network branded as community. Second, audit your loyalty program for Market-Pricing-spoofed-as-Communal-Sharing — discount tiers labeled with member-of-the-family copy. These two audits will surface most of the high-leverage Fiske violations in your product.

What’s the single best book or paper to read on Relational Models Theory?

Fiske’s 1992 Psychological Review paper “The Four Elementary Forms of Sociality” is the most compact authoritative source. His 1991 book Structures of Social Life is the long version with the ethnography. For applications across psychology, the 2004 edited volume by Nick Haslam, Relational Models Theory: A Contemporary Overview, is the best one-stop tour of how the theory has held up across the literatures it touches.

References

1. Fiske, A. P. (1991). Structures of Social Life: The Four Elementary Forms of Human Relations. Free Press.
2. Fiske, A. P. (1992). The four elementary forms of sociality: Framework for a unified theory of social relations. Psychological Review, 99(4), 689–723.
3. Fiske, A. P. (2004). Four modes of constituting relationships. In N. Haslam (Ed.), Relational Models Theory: A Contemporary Overview (pp. 61–146). Lawrence Erlbaum.
4. Haslam, N. (Ed.). (2004). Relational Models Theory: A Contemporary Overview. Lawrence Erlbaum.
5. Fiske, A. P., & Tetlock, P. E. (1997). Taboo trade-offs: Reactions to transactions that transgress the spheres of justice. Political Psychology, 18(2), 255–297.
6. Heyman, J., & Ariely, D. (2004). Effort for payment: A tale of two markets. Psychological Science, 15(11), 787–793.
7. Clark, M. S., & Mills, J. (1979). Interpersonal attraction in exchange and communal relationships. Journal of Personality and Social Psychology, 37(1), 12–24.
8. McGraw, A. P., & Tetlock, P. E. (2005). Taboo trade-offs, relational framing, and the acceptability of exchanges. Journal of Consumer Psychology, 15(1), 2–15.
9. Cropanzano, R., & Mitchell, M. S. (2005). Social exchange theory: An interdisciplinary review. Journal of Management, 31(6), 874–900.
10. Vodosek, M. (2009). The relationship between relational models and individualism and collectivism: Evidence from culturally diverse work groups. International Journal of Psychology, 44(2), 120–128.
11. Sahlins, M. (1972). Stone Age Economics. Aldine-Atherton.
12. Bird-David, N. (1990). The giving environment: Another perspective on the economic system of gatherer-hunters. Current Anthropology, 31(2), 189–196.
13. Tetlock, P. E. (2003). Thinking the unthinkable: Sacred values and taboo cognitions. Trends in Cognitive Sciences, 7(7), 320–324.
14. Bloch, M., & Parry, J. (Eds.). (1989). Money and the Morality of Exchange. Cambridge University Press.
15. Rai, T. S., & Fiske, A. P. (2011). Moral psychology is relationship regulation: Moral motives for unity, hierarchy, equality, and proportionality. Psychological Review, 118(1), 57–75.

Related Reading

• Social Identity Theory — the deeper account of how a Communal Sharing in-group gets formed and policed.
• Interdependence Theory — the formal outcome-matrix substrate under Equality Matching and Market Pricing dyads.
• The Investment Model of Commitment — why the four-model stack predicts retention better than any single one.
• The Octalysis Framework — the eight Core Drives that pair with the four models to form the full design grammar.
• The Behavioral Framework Library — every behavioral and motivation framework Yu-kai has written an S-Tier guide to.

WOULD YOU LIKE YU-KAI CHOU TO WORK WITH YOUR ORGANIZATION?