Skip to content
Primores AI wiki

TPB (Theory of Planned Behaviour) — What It Means

TPB (Theory of Planned Behaviour)

TL;DR: TPB explains why people intend to do things based on three factors: their personal attitude, what others think they should do (subjective norms), and whether they believe they can actually do it (perceived control). For AI adoption, surprisingly, peer pressure matters less than personal motivation.

Simple Explanation

The Theory of Planned Behaviour (TPB) is a psychology framework developed by Icek Ajzen in 1985 to predict and understand human behavior. It says that your intention to do something depends on three things:

  1. Attitude: Do you think it’s a good or bad idea?
  2. Subjective Norms: Do people important to you think you should do it?
  3. Perceived Behavioral Control: Do you believe you can actually do it?

When all three align positively, you’re much more likely to act.

The Three Components

1. Attitude

Your personal evaluation of the behavior — is using AI shopping tools good or bad for you?

In e-commerce AI context:

  • Do consumers trust that AI recommendations are helpful?
  • Do they believe AI will handle their data responsibly?
  • Do they see AI as making shopping easier or more complicated?

Key insight: Trust and faith in AI significantly affect attitude. Consumers who believe AI is reliable and in their interest have more positive attitudes.

2. Subjective Norms

Social pressure — what do people around you expect?

In e-commerce AI context:

  • Do friends and family use AI shopping features?
  • Do influencers recommend AI-powered platforms?
  • Is AI shopping socially accepted in their culture?

General finding: Across the 16-study synthesis in Ajzen (1991), subjective norms were the weakest of the three predictors for most behaviors — “personal considerations tended to overshadow the influence of perceived social pressure.” AI-adoption research is a clean confirmation: peer pressure has weak correlation with AI acceptance. The pattern is general; AI isn’t a special case, just an unusually unambiguous one.

3. Perceived Behavioral Control (PBC)

Your belief about your ability to perform the behavior — can you actually use AI effectively?

In e-commerce AI context:

  • Can consumers easily navigate AI recommendation systems?
  • Do they understand how to customize AI preferences?
  • Can they override AI decisions when needed?

Key insight: PBC has a direct positive effect on both perceived ease of use AND purchasing behavior. Consumers who feel in control engage more.

PBC sits in a small family of constructs that get conflated:

ConstructWhat it measuresDomain
Perceived behavioral control (Ajzen)Ease/difficulty of this specific behaviorBehavior-specific
Self-efficacy (Bandura)Confidence in executing courses of actionBehavior-specific (≈ PBC)
Locus of control (Rotter)Whether outcomes generally depend on you or external forcesGeneralized life-trait

Practical implication for marketers and product designers: measure “can the user do this?” not “does the user feel in control of their life?” Generalized control measures (Rotter’s scale) often fail to predict specific behaviors. PBC and self-efficacy don’t — because they’re scoped to the behavior at hand.

Ajzen explicitly aligns PBC with Bandura’s self-efficacy: “the present view of perceived behavioral control is most compatible with Bandura’s concept of perceived self-efficacy.” The two are essentially the same construct viewed from a behavior-prediction model versus an agency-mechanism one.

Empirical Performance

How well does TPB actually predict? Across the studies synthesized in Ajzen (1991):

What is predictedAverage multiple correlation (R)Range across studies
Behavior (from intention + PBC)~.51.20 to .78
Intention (from attitude + SN + PBC)~.71.43 to .94

Translation: roughly 25-50% of variance in actual behavior is captured by intention + PBC; roughly 50% of variance in intention is captured by the three TPB predictors. These are the canonical baseline numbers — useful when justifying the model in any application.

Where the model is weakest: behaviors with low volitional control (losing weight, getting an ‘A’ in a course). The mismatch between perceived and actual control caps prediction.

Where the model is strongest: behaviors with high volitional control (voting choice, simple consumer choices). There, intention alone reaches r=.75–.84.

A surprise from the 1991 synthesis: the expected intention × PBC interaction (motivation × ability multiplicatively) rarely emerges empirically. Linear additive models fit the data better. Practical implication: gains in either intention or PBC are roughly linear, not multiplicative.

Why It Matters for Business

For E-commerce

TPB helps explain why consumers adopt (or resist) AI features:

FactorHigh = More AdoptionLow = Resistance
Attitude”AI saves me time""AI is creepy/invasive”
Subjective Norms”Everyone uses it""My friends don’t trust it”
PBC”I can control it""It’s too complicated”

For Marketing Strategy

  1. Build positive attitudes through transparency and demonstrated value
  2. Don’t over-rely on social proof — personal benefits matter more for AI
  3. Maximize perceived control through user-friendly design and customization options

TPB vs. Other Models

ModelWhat It ExplainsBest For
TPBWhy people intend to actPredicting planned behavior, understanding resistance
TAMWhy people accept technologyTechnology adoption, feature design
S-O-RWhy people respond emotionallyImpulse behavior, experience design

How They Work Together

  • S-O-R captures emotional triggers (stimulus → feeling → response)
  • TAM captures rational assessment (usefulness + ease = acceptance)
  • TPB captures intentional factors (attitude + norms + control = intention)

A complete understanding of consumer behavior requires all three perspectives.

Key Research Findings

From systematic literature reviews on AI in e-commerce:

  1. Trust is foundational — Faith in AI significantly affects consumers’ perception and interaction with AI tools

  2. Subjective norms are often the weakest of the three predictors — Across the 16-study synthesis in Ajzen (1991), subjective norms had mixed or insignificant coefficients for most behaviors, with personal considerations overshadowing perceived social pressure. The finding generalizes; AI adoption is a clean confirmation, not a special case.

  3. PBC directly affects purchase behavior — Consumers who feel in control are more likely to buy

  4. Cultural context matters — In cultures where AI is encouraged and valued, consumers perceive AI tools as more useful and easier to use

  5. Ethics shape attitude — Consumer awareness of privacy implications can promote or hinder AI adoption

When the Basic Three-Factor Model Isn’t Enough

The 1991 paper is honest about where the basic model has limits. Three caveats worth carrying:

Past Behavior as a Residual Predictor

Even with intention and PBC controls, past behavior often retains a significant residual effect on later behavior. The model isn’t always sufficient — past behavior captures something the basic predictors miss (habit, unmeasured factors, or method variance shared by behavioral measures).

Marketing implication: stated intentions don’t fully predict; behavioral history adds signal that surveys miss. CRM data and behavioral logs have predictive validity beyond intention surveys, even when the surveys hit the right TPB constructs.

Moral Obligation as a Fourth Predictor

For ethical/unethical behaviors specifically, perceived moral obligation adds 3-6% explained variance beyond the basic three predictors (Beck & Ajzen, 1991). The model is extensible: domain-specific factors can be added when relevant.

Marketing implication: in regulated, sensitive, or ethics-adjacent categories (financial services, health, gambling, sustainability claims), expect attitude + subjective norm + PBC to under-predict. Moral and identity factors carry weight there.

Methodological Caveat — The Belief Layer Is Leakier Than It Looks

The expectancy-value formulation underneath TPB (the belief × evaluation products that build attitudes) only explains 10-36% of variance in global attitude measures, even with optimal rescaling. The model surface looks clean but the bottom layer has measurement leak.

Practical implication: belief-elicitation surveys are useful but partial — they don’t fully capture how attitudes form. When designing TPB-based research or persuasion strategies, treat belief inventories as one signal, not a complete picture.

Practical Applications

For Product Teams

  • Design AI features that feel controllable, not autonomous
  • Provide clear explanations of AI decisions
  • Allow easy customization of AI behavior
  • Build in override mechanisms

For Marketing Teams

  • Focus on personal benefits rather than social proof for AI features
  • Address privacy and ethical concerns proactively
  • Use education-focused content, not just endorsements
  • Consider cultural context in messaging

For Customer Experience

  • Offer tutorials for AI features
  • Provide visible preference controls
  • Show “why” behind AI recommendations
  • Make opt-out easy and frictionless

Key Takeaways

  • TPB explains intended behavior through attitude, social norms, and perceived control
  • Subjective norms are often the weakest of the three predictors generally — peer pressure matters less than personal motivation across many behaviors, AI adoption included
  • Perceived behavioral control directly affects purchase behavior
  • Trust and transparency are essential for positive attitude formation
  • Cultural context significantly affects AI acceptance rates

Sources

  • Ajzen, I. (1991). The Theory of Planned Behavior. Organizational Behavior and Human Decision Processes, 50, 179–211. — The foundational synthesis paper. Empirical baseline (16–19 studies), full theoretical treatment, methodological caveats. The primary citable source for any TPB claim.
  • Ajzen, I. (2014). “The Theory of Planned Behaviour is Alive and Well.” Health Psychology Review. — Defense paper; useful but downstream of the 1991 synthesis.
  • Beck, L., & Ajzen, I. (1991). “Predicting dishonest actions using the theory of planned behavior.” Journal of Research in Personality. — The moral-obligation extension (cheating, shoplifting, lying study).
  • Bandura, A. (1982). “Self-efficacy mechanism in human agency.” American Psychologist, 37, 122–147. — The self-efficacy concept Ajzen aligns PBC with.
  • Marshall, S. (2024). “A systematic analysis of AI in digital marketing and its effects on consumer behaviour and decision making in E-commerce.” University of Bedfordshire Dissertation. — Multi-framework synthesis applying TPB to AI commerce.
  • Lopes, J.M. et al. (2024). “AI Meets the Shopper: Psychosocial Factors in Ease of Use.” Behavioral Sciences. — TPB application to AI commerce.
  • Kaplan, H.E. (2018). “Factors Determining E-Consumer Behavior.” International Review of Management and Business Research. — TPB in e-commerce context.

Last updated: 2026-04-20