Interactive tools and engagement mechanisms

Interactive tools are software features on a small-business website that accept user input and return a computed or looked-up result. The category includes calculators, configurators, multi-step forms, quizzes, search interfaces, booking flows, and dashboards. Their distinguishing property is contingency: the output depends on what the user entered.

This page is a reference catalogue of the psychological mechanisms that explain why interactive tools deepen a visitor's relationship with a small business — the mechanisms that operate during the brief window in which a user types numbers, picks options, drags a slider, or watches a result update. It is the applied companion to Behavioural economics for small-business marketing, which surveys the underlying psychology in general terms.

The page is organised around eight independently evidenced mechanisms (curiosity, generation, active processing, agency and competence, personalisation, goal-gradient progress, the IKEA effect, and flow), followed by the binding constraints — variable-reward ethics, the inverted-U on curiosity, the "too much interactivity" cost, the limits of vendor outcome statistics, and the buyer-relationship shift that has made rep-free tools strategically valuable.

Defining the category

The defining test of an interactive tool is contingent output. A static page that says "call us" does not qualify; a quote form that returns a price range based on inputs does. The technical floor for building such tools at small-business scale has fallen sharply, which is why the category has moved from enterprise-only to ordinary small-business reach.

An interactive capability is one where the visitor supplies input and the site returns a computed or looked-up result — calculator, quote / estimate tool, search box, booking flow, configurator. The defining test is that the output depends on what the user entered.

Source: Industry-consensus definitional framing. Confidence: Industry-consensus (definitional).

Caveat: No clean current primary dataset ties interactive features specifically to conversion or retention for general small-business marketing websites; the case rests on definitional and mechanism grounds, not strong outcome data.

A second distinction, drawn from human-computer interaction theory, separates modality interactivity (sliders, drags, zooms — the user manipulates the interface) from message interactivity (the system responds contingently to user input — the defining feature of calculators and quizzes).

S. Shyam Sundar's Theory of Interactive Media Effects (TIME, 2015) distinguishes modality interactivity from message interactivity. Modality interactivity refers to slide / drag / zoom interactions. Message interactivity refers to a system that responds contingently to user input. The latter is the structurally distinct feature of calculators and quizzes.

Source: Sundar et al. (2015) TIME framework. Confidence: Verified for framework.

The mechanisms below operate primarily through message interactivity. Modality features are useful only insofar as they support the contingent exchange — a point that becomes important when the "too much interactivity" caveat is considered.

The eight-mechanism framework

The mechanism-level evidence for why interactive tools engage users converges on eight independent literatures, each of which has been peer-reviewed and meta-analysed in its own right. The convergence — not any single study — is what gives the framework its weight.

Strong independent evidence sits at the mechanism level — cognition, memory, motivation, communication and HCI lab and field studies. Curiosity, generation effect (d ≈ 0.40), active learning (+0.47 SD), choice and intrinsic motivation, agency in customisation, modality-interactivity absorption, and motivating-uncertainty all draw from multiple independent literatures. Mechanisms reinforce each other. The convergence is the robust core.

Source: Cross-literature synthesis, June 2026. Confidence: Verified (mechanism level).

Caveat: Independent evidence for specific business outcomes of interactive tools (conversion lifts, time-on-tool) is weak and largely vendor-sourced. The mechanism case is robust; the outcome case is not.

Across the brief, the business-outcome statistics ("interactive content gets 2× engagement," "52.6% more engagement," "4-5× more pageviews") almost all trace back to vendors that sell interactive-content platforms — ion interactive (now Rock Content), Demand Metric (vendor-sponsored), Mediafly, Outgrow, Ceros, SnapApp. The single most-quoted source (Demand Metric 2014) is explicitly sponsored by ion interactive, n=185 marketers, opinion survey, not behavioural research.

Source: Source-audit of interactive-content business statistics. Confidence: Verified (source audit).

Caveat: The practical posture is to lead on mechanism evidence and treat outcome statistics as marketing.

Each mechanism has its own boundary conditions. None is an unconditional "more is better" lever. The sections below catalogue each mechanism with its supporting evidence and its known limit.

Generation effect and active production

The generation effect is the finding that information produced by the learner is remembered better than information merely read. It is one of the most-replicated effects in cognitive psychology.

Bertsch, Pesta, Wiscott & McDaniel (2007), Memory & Cognition 35(2), 201-210 — 86-study meta-analysis, 445 effect sizes. Canonical effect size d ≈ 0.40 ("almost half a standard deviation"). The effect is larger at longer retention intervals: d ≈ 0.64 for >1 day vs. ~0.32 immediate.

Source: Bertsch et al. (2007), Memory & Cognition. Confidence: Verified (meta-analysis).

A second-generation meta-analysis a decade later refined the design dial: the size of the effect depends on how constrained the produced response is.

McCurdy et al. (2020), Psychonomic Bulletin & Review — meta-analysis of 126 articles, 310 experiments. Effect magnitude depends on "generation constraint" — how constrained the produced response is.

Source: McCurdy et al. (2020), Psychonomic Bulletin & Review. Confidence: Verified (meta-analysis).

Practical consequence: an input that the user types or actively configures produces more durable memory than the same answer selected from a dropdown of canned options. The same content is "owned" differently depending on whether the user generated it or recognised it.

A connected and broader-scoped literature is the active-learning meta-analysis from education research.

Freeman et al. (2014), PNAS 111(23), 8410-8415 — 225-study meta-analysis of active learning. Exam performance +0.47 SD under active learning; odds of failing 1.95× higher under passive lecturing. Robust to publication-bias checks.

Source: Freeman et al. (2014), PNAS. Confidence: Verified (meta-analysis, top journal).

Caveat: Effect is largest in small classes and dependent on volunteer instructors; generalisability if universally mandated is described by Freeman et al. as "an open question." Education-context evidence; the bridge to commercial tools is conceptual, not measured.

The conceptual scaffolding for the active-versus-passive distinction is the ICAP framework, which adds two finer-grained intermediate categories.

Chi & Wylie (2014), "The ICAP Framework," Educational Psychologist 49(4), 219-243. Engagement hierarchy: Interactive > Constructive > Active > Passive. The authors report ~8-10% learning improvement with each step up the engagement hierarchy.

Source: Chi & Wylie (2014), Educational Psychologist. Confidence: Verified for the framework.

Caveat: ICAP's behaviour-based coding is acknowledged as a limitation — overt behaviour is not a guarantee of cognitive engagement.

The generation effect has known ceilings. It is robust at the word level and at moderate text lengths; it ceases to produce reliable gains for long expository text. A tool that asks the user to type a phone number, a square-footage figure, or a project description benefits; a tool that asks the user to "write a paragraph describing your business" does not.

A practical formulation of the design dial:

Where appropriate, make the user generate inputs (free-text where useful, short typed answers, configured-by-the-user options) rather than picking from canned dropdowns. The generation effect ceilings beyond ~900 words and does not reliably scale to expository text. Default to short typed inputs over dropdowns when both are equally usable. Do not over-apply: a request to "generate a paragraph describing your business" defeats the mechanism.

Source: Rule synthesised from Slamecka & Graf 1978, Bertsch 2007, McCurdy 2020, and the 2023 expository-text ceiling literature. Confidence: Verified.

IKEA effect and effort-based valuation

The IKEA effect is the phenomenon in which people place a higher value on products they have helped build. In the context of interactive tools, the relevant artifact is a completed configuration — a quote, a saved plan, a chosen specification.

The mechanism has a binding precondition: completion. Building-then-destroying or failing to complete eliminates the effect. An abandoned configurator does not produce ownership; it produces friction.

Pair the IKEA effect with completion. Only completed configurations produce psychological ownership. Building-then-destroying or failing eliminated the effect in the original Norton-Mochon-Ariely (2012) experiments. An abandoned configurator produces no ownership boost and may actively frustrate.

Source: Norton-Mochon-Ariely 2012, Sarstedt et al. 2017 conceptual replication. Confidence: Verified (with replication).

Practical consequence: default toward fewer steps and save-state support. A user who completes a three-step configuration owns the result; a user who abandons a twelve-step configuration owns nothing and remembers the friction. Save-and-share patterns turn the configuration into a named artifact and reinforce ownership across sessions.

The IKEA effect pairs naturally with goal-gradient design (see below) and with the agency literature: both push toward making finishing easy, visible, and attributable to the user's effort.

Active processing through customisation and agency

The HCI literature distinguishes two superficially similar interaction patterns. Customisation is when the user acts — the user selects, configures, or adjusts. Personalisation is when the system acts — the system tailors based on inferred or stored attributes. The engagement and attitudinal evidence prefers customisation specifically because of the agency it confers.

Sundar & Marathe (2010), "Personalization versus Customization," Human Communication Research 36(3), 298-322. User-tailored ("customization," where the user acts) versus system-tailored ("personalization," where the system acts). The appeal of customisation is tied to the user's sense of agency; "power users" rated customisable interfaces higher.

Source: Sundar & Marathe (2010), HCR. Confidence: Verified — directly relevant HCI evidence.

This isolates doing it yourself as the engaging element, bridging self-determination theory directly into tool-design terms.

Adjacent to agency is competence, and the most-replicated competence-related finding is that positive contextual feedback raises intrinsic motivation through perceived competence.

Vallerand & Reid (1984), Journal of Sport Psychology 6, 94-102 (N=115 phase 1; 84 returned). Positive feedback increased while negative feedback decreased both intrinsic motivation and perceived competence; path analysis supported the mediating effects of perceived competence on intrinsic motivation.

Source: Vallerand & Reid (1984), Journal of Sport Psychology. Confidence: Verified.

In tool terms, a clean input that produces a clear positive confirmation ("got it," a check mark, an updated total) multiplies engagement; an unclear or error-pronouncing tool inverts the mechanism.

Choice itself supports agency, but with a sharply bounded effect size.

Patall, Cooper & Robinson (2008), Psychological Bulletin 134(2), 270-300 — 41-study meta-analysis. Providing choice enhanced intrinsic motivation, effort, task performance, and perceived competence.

Source: Patall, Cooper & Robinson (2008), Psychological Bulletin. Confidence: Verified (meta-analysis).

Caveat: Effect was modest and moderated — stronger for instructionally irrelevant choices, when 2-4 successive choices were given, when no extrinsic reward followed, and for children more than adults. Too much choice or choice that adds burden can wash out the benefit.

The practical synthesis is to support autonomy and competence without crossing into choice overload.

Design tools to support autonomy and competence: 2-4 meaningful choices per step (not 12); positive contextual feedback when inputs succeed; let the user act rather than auto-populating everything. Prefer user-supplied inputs over system-inferred defaults when the user can supply them without friction.

Source: Synthesis of Deci-Ryan SDT, Patall et al. 2008, Sundar-Marathe 2010, Vallerand-Reid 1984. Confidence: Verified.

Self-reference and personalisation in tool outputs

The strongest single-mechanism evidence in the cluster is for self-reference — information related to the self is processed more deeply and remembered better. In the tool context, the relevant lever is whether the result is the user's own number, computed from the user's own inputs, or a generic placeholder.

Maximise self-relevance in the output. Surface the user's own number computed from the user's own inputs, not a generic "average customer like you sees X" framing. Personalisation is the best-evidenced mechanism in the brief — Symons-Johnson 1997 meta on self-reference, Svensson 2022 on self-relevance and attention, Tam-Ho 2006 on web personalisation mediation, De Keyzer 2025 meta on personalised advertising — actual personalisation (computed from real input) outperforms hypothetical or scenario personalisation via stronger self-referencing.

Source: Synthesis of Symons-Johnson 1997, Svensson 2022, Tam-Ho 2006, De Keyzer 2025. Confidence: Verified.

Practical consequence: result pages lead with "Your estimate: $X," not "Customers your size typically see $X." When confidence intervals are warranted, the user's number is anchored first, the range second. Hypothetical or second-person-plural framings dilute the self-reference advantage and should be avoided.

Self-reference also has a memory side-effect: a curious user encodes the surrounding content (brand name, context, sales positioning) more strongly than a non-curious one. The tool not only delivers its answer; the surrounding brand context sticks.

Curiosity gaps and the information-deprivation mechanism

Curiosity is treated in this literature as a state in which an information gap motivates seeking, and the neural correlates of curiosity overlap with reward processing. Two foundational studies anchor the modern case.

Kang, Camerer, Loewenstein et al. (2009), "The Wick in the Candle of Learning," Psychological Science 20(8), 963-973. fMRI while reading trivia questions: curiosity correlated with activity in caudate regions tied to anticipated reward. In a behavioural study, people spent scarce resources (limited tokens or waiting time) to learn answers when more curious. Curiosity predicted better recall of surprising answers 1-2 weeks later. Pupil dilation tracked curiosity.

Source: Kang et al. (2009), Psychological Science. Confidence: Verified.

Caveat: Reward and learning study, not an engagement-with-tools study; the bridge to interactive tools is inferential.

Gruber, Gelman & Ranganath (2014), "States of Curiosity Modulate Hippocampus-Dependent Learning via the Dopaminergic Circuit," Neuron 84(2), 486-496. fMRI: high-curiosity states enhanced midbrain (SN/VTA) and nucleus accumbens activity and improved memory both for the curiosity-target information and for incidental material encountered during the curious state, on immediate and one-day-delayed tests.

Source: Gruber et al. (2014), Neuron. Confidence: Verified.

The "people spent tokens to learn" finding generalises cleanly to "users will give some friction — form fields, time — to satisfy curiosity a tool has opened." But this lever has two important limits.

The first limit is the inverted U over prior knowledge. Curiosity peaks at moderate knowing, not at total ignorance and not at familiar territory.

Curiosity follows an inverted-U over prior knowledge / confidence. Curiosity peaks at moderate prior knowledge and falls when one knows nearly nothing or nearly everything. Replicated across paradigms (trivia, blurred pictures, letter guessing, web-ad clicks).

Source: Kang et al. 2009; Dubey & Griffiths (2020), Psychological Review 127(3), 455-476; Lee et al. (2024), Metacognition and Learning. Confidence: Verified.

The second limit is that overly vague teasers backfire — they reduce, not increase, information-seeking.

Curiosity gaps can backfire. When a teaser is too vague or abstract, information-seeking drops (Scientific Reports 2024, "When curiosity gaps backfire"). Information gaps also produce frustration, not just pleasant curiosity, in some contexts (Organizational Behavior and Human Decision Processes 2023).

Source: Scientific Reports 2024; OBHDP 2023. Confidence: Verified (limit / boundary condition).

A tool's framing question must therefore be concrete enough that the user can picture a satisfying answer at the other end and aimed at a level of knowledge between "I have no idea" and "I already know."

Frame the tool's opening question so it feels answerable-but-unknown to the user. Aim above the "I have no idea" floor and below the "I already know" ceiling. Tool titles should name the specific question the user is here to answer ("What's the cost of a new roof for a 2,000 sq ft home?"), not abstract teasers ("Discover your savings"). Avoid "you won't believe what you'll save" framings — too-vague curiosity is documented to backfire.

Source: Synthesis of Loewenstein 1994, curiosity inverted-U literature, Scientific Reports 2024. Confidence: Verified.

Goal-gradient and progress-toward-reward

The goal gradient is the finding that motivation accelerates as a structured task nears completion. The most-cited applied demonstration is the Kivetz coffee-card field study, in which pre-filling two of twelve stamps shortened the time to completion of a card with the same effective steps.

The tool-design implications are straightforward: every multi-step tool needs visible progress; per-step interaction cost should be low; and the starting state should not be empty when an honest pre-fill is available.

Design every multi-step tool for the goal gradient: visible step counter or progress bar; low per-step interaction cost; non-empty starting state when honestly possible (pre-fill common defaults, the way the Kivetz coffee card pre-filled two of twelve stamps). "Step 3 of 5" beats "Continue." Honest pre-fills are better than blank fields. For tools that cannot be completed in one sitting, pair with a save-and-resume pattern.

Source: Kivetz et al. 2006; Nielsen Norman Group engagement model. Confidence: Verified.

The engagement-as-utility frame from Nielsen Norman Group operationalises the progress mechanism in interface terms.

Nielsen Norman Group frames user engagement as expected utility = perceived value minus interaction cost. A well-built tool delivers a high-value answer for modest, well-signposted effort, raising expected utility. Abandonment can happen within seconds when perceived value drops.

Source: nngroup.com engagement and form-design literature. Confidence: Industry-consensus (NN/g is an independent UX authority with no portal or calculator vendor incentive).

A tool that fails the value-minus-cost test is abandoned regardless of how well its psychology is grounded.

Flow and proximal-goal design

Flow — the state of absorption in an optimally challenging task — is a frequently invoked but empirically uneven anchor for tool design. The honest reading of the evidence is that two of flow's three classical antecedents (clear goals; immediate feedback) are robust and engineerable, while the third (challenge-skill balance) is moderate and contested.

Fong, Zaleski & Leach (2015), "The challenge-skill balance and antecedents of flow," Journal of Positive Psychology 10(5), 425-446 (28 studies meta). The challenge-skill balance to flow relationship was "moderate." Compared to other theorised antecedents, challenge-skill balance was a robust contributor along with clear goals and sense of control.

Source: Fong, Zaleski & Leach (2015), JPP. Confidence: Verified (meta-analysis).

The practical posture for short tool sessions is to engineer the robust components and to make modest, accurate claims about absorption rather than deep-flow.

Engineer the robust flow components: clear proximal goal ("get your estimate") and immediate feedback (output updates as inputs change). Do not promise clients "deep flow" for short tool sessions, and do not anchor recommendations on the challenge-skill balance — it is moderate and contested. In client copy, frame as "supports user absorption," not "puts users in flow state." Result page updates live as the user changes inputs. Single clear goal per tool; don't bury the goal in feature framing.

Source: Synthesis of Csikszentmihalyi 1990, Fong-Zaleski-Leach 2015 meta, 2025 systematic review on flow measurement. Confidence: Verified.

The "immediate feedback" dimension reinforces a separate finding from the HCI absorption literature.

Oh & Sundar (2015), "How Does Interactivity Persuade?," Journal of Communication 65(2), 213-236 — N=167 factorial experiment. Modality interactivity (slider) produced more positive interface assessment, greater cognitive absorption, and more favourable attitudes.

Source: Oh & Sundar (2015), JoC. Confidence: Verified.

That finding is, importantly, paired with a counter-finding from the same paper that sets the boundary for the next section.

Variable-ratio reinforcement and uncertainty incentives

A recurring temptation in interactive-tool design is to import variable-ratio reinforcement — the engagement engine behind slot machines and social-media feeds — into ordinary commercial tools. The schedule-of-reinforcement literature is dominantly a gambling and compulsion literature; citing Skinner-box research for benign tool engagement is a stretch that the underlying evidence base does not support.

The clean benign literature on motivating uncertainty (Shen, Fishbach & Hsee, JCR) does demonstrate that uncertainty about a reward can motivate effort — but only under two specific conditions: the user is process-focused, not outcome-focused, and the uncertainty resolves immediately. Under outcome focus, the effect reverses to uncertainty aversion.

When variable or uncertain feedback is used in a tool design, cite the benign motivating-uncertainty literature (Shen-Fishbach-Hsee 2015; Shen-Hsee-Talloen 2019) — not Ferster & Skinner or Schultz primate electrophysiology. Respect the process-focus and immediate-resolution conditions. Treat the mechanism as conditional and ethics-flagged.

Source: Synthesis of motivating-uncertainty literature versus schedules-of-reinforcement literature. Confidence: Verified.

Caveat: The classic VR-schedule literature is dominantly a gambling and compulsion literature. Use variable elements sparingly and ethically; frame the user's focus on process ("explore your options"), not outcome ("win the bigger prize"); resolve uncertainty immediately. If a design starts looking slot-machine-shaped, kill it.

This is the dark-pattern boundary for the cluster. The honest case for interactive tools rests on the eight mechanisms catalogued in the preceding sections; uncertainty schedules are not part of that case unless engineered narrowly within the Shen-Fishbach-Hsee conditions.

The "too much interactivity" limit

The most under-appreciated finding in the literature is that interactivity has a real cost in deep processing. The same Oh & Sundar paper that documents absorption gains also documents a reduction in message-related thoughts.

Interactivity is not uniformly positive. Oh & Sundar 2015 found modality interactivity reduced the number of message-related thoughts — absorption can come at the cost of deep elaboration. Sundar's broader work warns of a "too much interactivity" cost: added interactive features can overload and reduce processing or attitudes for some users and tasks.

Source: Oh & Sundar (2015), JoC (same paper); Sundar broader work. Confidence: Verified (limit).

The practical formulation is restraint.

Add interactive features only where they let the user accomplish something necessary to the tool's purpose. Decorative interactivity (sliders, animations, drag-zooms for their own sake) reduces deep elaboration and can net-decrease engagement quality. Every interactive feature must answer "what user task does this enable?" If the answer is decorative, remove it. Prefer message interactivity (the system responds contingently to user input — calculator core) over modality interactivity (sliders, zooms, drags) when they compete for screen space. When stakes are high — the user needs to understand the output, not just see it — reduce interactivity around the result to support elaboration.

Source: Synthesis of Oh-Sundar 2015 counter-finding, Sundar broader work, ICAP framework limitations. Confidence: Verified.

The buyer-relationship shift

A separate strand of evidence — independent of the cognitive and HCI mechanisms above — has reshaped the strategic position of interactive tools in B2B and professional-services contexts. The modern B2B buyer prefers to find answers themselves.

Gartner (March 2026 survey of 646 B2B buyers, Aug-Sep 2025): 67% of B2B buyers prefer a rep-free experience; 45% used AI during a recent purchase.

Source: Gartner press release / report, March 2026. Confidence: Verified / Industry-consensus (Gartner is an independent research firm).

Caveat: B2B-specific. Applying it to all SMB audiences is a reasonable directional extension but not a measured fact for, say, B2C local services.

The earlier Gartner survey adds a complementary finding that is directly actionable: when a vendor's website contradicts the rep, buyers walk.

Earlier Gartner survey of 632 buyers, Aug-Sep 2024: 61% prefer an overall rep-free buying experience; 73% actively avoid suppliers who send irrelevant outreach; 69% report inconsistencies between a vendor's website and what reps tell them.

Source: Gartner, Aug-Sep 2024 buyer survey. Confidence: Verified.

For interactive tools, this produces two design implications. The first is placement: the tool sits where the rep-free buyer wants to be met.

For B2B and professional-services SMB clients, place the interactive tool before the contact form in the user journey, not behind it. The modern B2B buyer wants to find the answer, not be sold it. Default tool placement: result page first; CTA to talk to a human is the secondary action, available but not gating. For B2C and local-services contexts, the directional logic still holds, but the Gartner numbers are B2B-specific — present as direction, not specific magnitude.

Source: Synthesis of Gartner 2024 and 2026 buyer surveys. Confidence: Verified for B2B.

The second is integrity: the tool's number must be one sales can meet or honestly explain.

Before any customer-facing tool ships, confirm that sales can meet or honestly explain the number the tool will display. The tool's output is the buyer's anchor — anchoring is robust even with experts (Northcraft-Neale 1987 real-estate experts) and even with judges (Englich-Mussweiler-Strack 2006 judges-and-dice). Gartner's 69% website-vs-rep inconsistency finding is the warning of what happens when on-site numbers and rep numbers diverge — buyers avoid those suppliers. A bait-and-switch tool destroys the trust the mechanism case earns.

Source: Synthesis of anchoring literature and Gartner 2024 inconsistency finding. Confidence: Verified.

Caveat: Pre-launch coordination between tool team and sales team is non-optional. If sales cannot meet the number for some segment, the tool's scope must exclude that segment or the tool must explicitly mark its output as "range / depends on assessment."

Signalling: brand innovativeness and word-of-mouth

A peripheral but on-topic mechanism is that the very fact of fielding a working interactive tool signals capability. The mediation evidence is single-source but peer-reviewed.

Pham et al. (2024), Australasian Marketing Journal, found brand innovativeness has a positive indirect effect on positive word-of-mouth, mediated by perceived brand expertise. A demonstrable capability — "we built something that works" — signals "we have the competence to deliver what we promised."

Source: Pham et al. (2024), AMJ. Confidence: Single-source (peer-reviewed).

Caveat: Independent research measures innovativeness → expertise → WOM generally, not calculators specifically. The link from "a calculator" to "perceived expertise" is a reasonable application of the theory, not a directly measured finding. A broken, inaccurate, or trivial tool signals the opposite (negative expertise); the rule is conservative — only ship tools that genuinely work.

Caveats and the boundary of the case

Several caveats run across the entire framework and are worth restating explicitly.

The evidence base is mechanism-level, not outcome-level. No clean, current, primary dataset ties interactive features specifically to conversion or retention for general small-business marketing websites.

No clean, current, primary dataset ties interactive features specifically (calculators, booking, account state) to conversion or retention for general SMB marketing websites. Available figures are vendor-sourced and frequently recycle the quarantined Mediafly statistic. The interactivity and account-state capabilities are argued on definitional and mechanism grounds, not on a strong outcome dataset.

Source: Literature search, June 2026. Confidence: Verified (gap-in-evidence).

Every mechanism is moderated. None is an unconditional lever:

• Curiosity follows an inverted-U over prior knowledge and backfires when teasers are too vague.
• The generation effect ceilings beyond ~900 words and does not reliably scale to expository text.
• Flow's challenge-skill balance is moderate and contested in measurement.
• Choice supports motivation at 2-4 options and washes out at more.
• Modality interactivity reduces message-related thoughts (the "too much interactivity" cost).
• Variable-ratio uncertainty motivates only under process focus with immediate resolution and otherwise reverses to aversion.

The dark-pattern boundary is real. The schedule-of-reinforcement literature is a gambling-and-compulsion literature, and importing it into commercial tools to manufacture engagement is both empirically unsupported and ethically suspect.

The buyer-relationship finding is B2B-specific. The Gartner rep-free numbers (67%, 61%, 69%) are robust for B2B contexts and directional but not measured for B2C local services.

Taken together, the case for interactive tools as engagement mechanisms is grounded but bounded: each mechanism is real and meta-analytically supported, the convergence across literatures is the source of the framework's weight, and every individual lever has known limits that constrain how aggressively it can be claimed. The underlying psychology — and the broader frame within which these mechanisms sit — is covered in Behavioural economics for small-business marketing; the related strategic question of why small-business owners often resist the marketing implications of these mechanisms is covered in Psychology of contractor marketing aversion.