Source-linked AI summary
Redesigning Trust: Replacing Dark Patterns with Fair Choice Architecture in Financial Interfaces
Oluwadamilola Awakan, Tawan Aroonwechkul, Roshan Gunjoor, Supriya Khadka, Sanchari Das
TL;DR
Digital financial platforms can make cancellation harder than enrollment, and existing detection and regulatory approaches do not give designers a constructive criterion for avoiding manipulation. The paper models provider-controlled effort as interaction cost and proposes a fairness constraint; its key result is that the verdict can be reached by counting interface components rather than resolving their effort weights.
Problem
Financial platforms can make cancellation more effortful than enrollment, while existing work detects such asymmetry after deployment without giving designers a constructive design criterion.
Source: §1 · pp. 1–2Method
The paper models interaction cost through navigation steps, mandatory inputs, and confirmation prompts, and requires offboarding cost not to exceed onboarding cost.
Source: §1 · p. 2, §3.2 · p. 4Results
The fairness verdict is independent of effort weights, so verification can rely on counting the interaction components rather than estimating their relative effort.
Source: §3.2 · p. 5Takeaways & Limitations
Designers can build toward an interaction-cost budget that auditors and regulators can test on an implemented interface.
Source: §7 · p. 10The work provides no evidence of user behavior and is instantiated for credit card enrollment and cancellation, with verification by inspection rather than user-outcome measures.
Source: §6 · p. 10
Abstract
from arXiv · showhide
Digital financial platforms make enrollment effortless and cancellation laborious. This asymmetry is a dark pattern that manipulates users who have already decided to leave. Existing work identifies such patterns after deployment, and regulators sanction them after harm, yet neither provides designers with a criterion for building interfaces that avoid manipulation. We model the provider as an adversary whose instrument is effort and express fairness as a constraint requiring that leaving never cost more than joining. Defining interaction cost over navigation steps, mandatory inputs, and confirmation prompts, we prove that this constraint holds for every assignment of effort weights if and only if no component of the exit flow exceeds its counterpart at entry. Fairness is therefore verifiable by counting rather than by estimating cognitive effort, and exact equivalence is unnecessary because exit legitimately requires fewer inputs than entry. We instantiate the model, together with invariants for visual parity and linguistic neutrality, in a mobile credit card prototype with parallel sign-up and cancellation workflows.
1 Introduction
Financial platforms can make cancellation harder than enrollment, using effort to discourage users who have decided to leave. The paper proposes an auditable fairness constraint and instantiates it in a mobile credit card prototype.
1 Introduction
- Cancellation friction can discourage users who intend to leave and expose them to ongoing financial obligations that conflict with their original choices.The paper frames this manipulation as a human-centered security challenge and argues for prevention during design, not only post-deployment detection.Source: §1 · pp. 1–2
- The paper proposes a fairness requirement that leaving must never cost more than joining, making fairness a measurable design invariant.The interaction-cost formulation counts navigation steps, mandatory inputs, and confirmation prompts; its verdict is independent of effort weights.Source: §1 · p. 2
- A mobile credit card prototype uses parallel sign-up and cancellation workflows to demonstrate structural, visual, and linguistic symmetry as auditable design invariants.The prototype is checked against the interaction-cost constraint by structured inspection.Source: §1 · p. 2
2 Related Work
Prior work describes and detects cancellation friction, while regulation has addressed the problem without operationalizing what makes cancellation as simple as enrollment. This paper turns symmetry into a design-time constraint with a verdict that does not depend on effort weights.
2 Related Work
- Dark-pattern taxonomies and empirical studies characterize interface mechanics and prevalence, while asymmetry and sludge describe provider-favored friction, including buried subscription cancellations.The cited literature treats cancellation friction as a low-level interface pattern that can be detected.Source: §2 · p. 2
- Regulatory action has targeted manipulative cancellation practices, but the paper says the principle of cancellation being as simple as enrollment remains operationally unspecified.The passage notes the 2025 vacatur of the 2024 rule on procedural grounds, without a ruling on substantive challenges.Source: §2 · p. 3
- Prior research describes asymmetry and audits deployed workflows, but does not provide designers a constructive criterion for avoiding manipulation during design.The paper distinguishes its design-time constraint from post hoc click-count comparisons: its compliance verdict is invariant to effort weights.Source: §2 · p. 3
3 The Symmetric Interaction Model
The model treats provider-controlled exit friction as an adversarial effort problem and makes fairness a componentwise constraint. It pairs that test with presentation invariants and an illustrative financial-service workflow.
3.1 Threat Model
- The provider is modeled as retaining a user already intent on leaving by adding effort to exit, while cancellation itself is ultimately completed.The model assumes the user has already formed the intention to cancel; shaping whether that intention forms is outside its scope.Source: §3.1 · pp. 3–4
3.2 Quantifying Interaction Cost
- Interaction cost sums weighted navigation transitions, mandatory inputs, and prompts; flow counts start at the authenticated home screen, with compound burdens categorized separately.A multi-field screen counts as one step, while fields are counted as inputs; prompts include confirmations, retention offers, and secondary warnings.Source: §3.2 · p. 4
- For every non-negative weighting, exit cost is no greater than entry exactly when each exit component is no greater; structural symmetry plus fewer inputs suffices.The proof establishes necessity by assigning all weight to any component that is larger in exit, and sufficiency because componentwise differences are non-negative.Source: §3.2 · p. 5
- The constraint rules out only componentwise asymmetry, not manipulation through language or timing, nor uniformly burdensome paired flows; absolute ceilings address the latter separately.The authors characterize the constraint as an auditable floor, not a sufficient account of fair choice architecture.Source: §3.2 · p. 6
3.3 Worked Example
- Three extra steps and two prompts illustrate inflated exit friction; the proposed cap uses three steps and one prompt, with zero exit inputs.The comparison is illustrative rather than measured from a particular interface; the proposed workflow satisfies cost non-inflation for every non-negative weighting.Source: §3.3 · p. 6
3.4 Design Invariants
- Visual parity requires equivalent styling, size, and prominence for primary actions, while linguistic neutrality requires factual, non-directive text.The invariants target presentation channels that step and prompt counts alone do not constrain.Source: §3.4 · p. 7
4 Prototype Implementation
The authors implemented parallel enrollment and cancellation workflows in a high-fidelity mobile credit card prototype. Inspection confirmed that the prototype met its structural, visual, and linguistic specifications, but did not test user behavior.
4 Prototype Implementation
- The prototype simulates account states such as outstanding balances and accrued reward points without live backend integration.The interface properties under examination do not require a live backend.Source: §4 · p. 7
- Enrollment and cancellation each use three mandatory transitions and one confirmation prompt, while enrollment requires four inputs and cancellation none.The optional cancellation feedback screen is excluded from the step count; the design adopts zero cancellation inputs even though the model permits fewer inputs than enrollment.Source: §4 · pp. 7, 9
- Primary actions receive equivalent styling, spacing, and placement; instructional and confirmation text is factual and non-directive, with no option preselected.These presentation-layer invariants constrain visual parity and linguistic neutrality.Source: §4 · p. 9
- Inspection confirmed structural symmetry and the absence of subordinate visual treatment or directive phrasing, but provided no evidence about user behavior.One author traversed both workflows and recorded the three interaction quantities per screen.Source: §4 · p. 9
5 Discussion and Implications
The model turns fairness into a countable design constraint that designers and auditors can apply. Its implications include an operational compliance test and component-level checks, while preserving room for legally required inputs.
5 Discussion and Implications
- Component-level invariants could support automated checks for structural asymmetry and visual coercion before deployment, though no checker was built.Designers and auditors can use the same counting procedure to assess whether a prompt fits within the exit budget.Source: §5 · p. 9
- The constraint operationalizes the “as simple as” principle with a compliance verdict that does not require calibrating the relative burden of dialogues and text fields.The authors note that a counting-only test also has estimable compliance costs, relevant to the cost-benefit analysis missing from the vacated rulemaking.Source: §5 · p. 9
- Because the constraint caps steps and prompts while allowing inputs to fall, providers can meet enrollment requirements and still offer an exit costing no more than entry.A genuinely mandated cancellation step would be visible as a deliberate exception rather than hidden among discretionary friction.Source: §5 · pp. 9–10
6 Limitations and Future Work
The work presents a formal constraint and prototype, but provides no evidence about user behavior. Its scope is limited to credit card workflows, and future work must test the effort measures and their use across settings.
6 Limitations and Future Work
- The study provides no evidence of user behavior, and its model and prototype cover only credit card enrollment and cancellation.Verification was by inspection against the specification, not by measuring completion time, error rate, or user trust.Source: §6 · p. 10
- Other services may involve workflow structures absent from the model, and applying the constraint across domains and jurisdictions remains future work.The authors cite multi-party contracts and regulatory verification steps as examples of structures not captured by the three quantities.Source: §6 · p. 10
- The three counted quantities may not capture users’ experienced effort, and studies should test whether visual parity and linguistic neutrality remain perceptually distinct from structural symmetry.The constraint is described as a necessary floor on one form of manipulation, not a validated model of user effort.Source: §6 · p. 10
7 Conclusion
The conclusion frames asymmetric cancellation friction as manipulation of users who have already decided to leave. It proposes a count-based fairness constraint whose verdict does not depend on effort weights.
7 Conclusion
- The constraint requires that leaving never cost more than joining, without requiring exact cost equivalence.The conclusion identifies steps, inputs, and prompts as the counted interaction components.Source: §7 · p. 10
- Because the verdict depends on counting rather than effort weights, auditors and regulators can evaluate an implemented interface without resolving the burden of each obstacle.The authors characterize the result as a budget rather than a debate.Source: §7 · p. 10