The UK’s Competition and Markets Authority (CMA) is the latest regulator to pile into online choice architecture (OCA). While its views on the ‘good and the bad’ of OCA echo views of other regulators, such as the ACCC, the CMA’s ‘taxonomy’ of OCA practices is a novel addition to the debate.
What is OCA?
OCA refers to the design of digital platforms, including the presentation and placement of choices and the design of interfaces. When conducted with the interests of consumers in mind, OCA can be incredibly beneficial; for example, it can recommend relevant products, and facilitate quick and seamless transactions and returns processes.
However, OCA can also be deployed in ways that harm consumers (for instance, relevant information can be hidden, default choices may not align with consumers’ preferences, and consumers’ attention and data may be exploited). These strategies, designed to take advantage of inherent psychological biases and lead users towards making certain choices, are known as dark patterns.
Cataloguing and grading the bad OCA
The CMA grouped 21 OCA practices into three broad categories:
- Choice structure (how choices are presented to consumers);
- Choice information (the information provided to consumers when presenting choices); and
- Choice pressure (how consumers' choices may be indirectly influenced).
Mimicking an online rating system (a touch of irony?), the CMA also assessed the strength of evidence for the ‘badness’ of an OCA practice using a star system: 1-star evidence rating (Theoretical) means that a concept has been discussed but lacks empirical validation; 2-star evidence rating (Empirical) means that a concept has been validated but lacks more robust data; 3-star evidence rating (Applicable) means that results are taken from controlled, reasonably powered trials; 4-star evidence rating (Replicable) means that the results have been successfully replicated in terms of setting, procedure and measurement; and 5-star evidence rating (Impact) means that result insights have been implemented and applied at scale.
Defaults got the 4 star ‘bad’ rating: ‘[d]efaults are one of the strongest and most reliable practices that influence consumers’. The CMA’s ‘comprehensive statistical analysis… of 58 academic studies into defaults found that a pre-selected default option is on average 27% more likely to be selected out of two options than if there were no default option'.
Why does the CMA say defaults are so bad?
‘First, they require less effort than making an active choice. This means that consumers who are in a hurry, not interested, or who have other demands on their cognition are more likely to stick with a default than to change it. Second, a default might imply endorsement or a recommendation by the choice architect, or that most consumers have chosen it. Finally, defaults may lead consumers to act as if they have already chosen the default option (called the ‘endowment effect’) and, consequently, they use the default as a reference point to construct their preferences. In fact, defaults are so powerful that, even when consumers are told they are about to be defaulted to a random choice, they can strongly influence important decisions.’
Choice architects can also choose what information they provide to consumers when presenting choices, frame the information in ways that highlight certain aspects over others, or make it harder to understand or find information.
As the stars in the right column indicate, there is a significant body of evidence around the OCA practice of ‘price dripping’. Price dripping occurs when the choice architect partially hides the price of a product or service, and only reveals the full price at later stages of the transaction.
However, the table shows that there is less evidence available for the practice of ‘information overload’. ‘Information overload’ occurs when the choice architect gives the consumer too much information about a product or service, thereby obscuring its most relevant features.
Choice architects can exert pressure on consumers to make certain choices using indirectly related factors listed in the table below:
The CMA acknowledges there is less evidence overall of the adverse effects of choice pressure. But ‘scarcity’ still ranks three stars of ‘badness’. The CMA says that while there is mixed evidence for whether supply-framed claims (e.g., “only 2 left!”) or demand framed claims (e.g., “25 customers are looking at this product”) are more effective, both show strong effects when applied by businesses to consumers.
Did they miss something?
Overall, the Discussion Paper – and particularly the taxonomy of OCA practices – is a helpful resource for competition and technology experts. Future analysis, however, would benefit from more nuance and recognition of the value judgements being made.
Is too much ‘science’ obscuring qualitative judgements?
As the Discussion Paper notes, OCA is a broad and neutral term – it can be used to further consumers interests or exploit them.
The CMA acknowledges that the same practice can be beneficial in some circumstances and not in others, or valued by some customers but harmful to others. For example, the OCA practice of ‘prompts and reminders’ involves the choice architect contacting the consumer to induce an action or follow up on a previous interaction. Although this can be harmful in some circumstances, some consumers may find it helpful. Similarly, ‘personalisation’ practices (where the choice architect uses data to personalise offers), may be perceived as invasive and pushy to some consumers, but useful and efficient to others.
These examples demonstrate that there are real difficulties in classifying OCA practices as either ‘good’ or ‘bad’. There are qualitative judgements being made here. As useful as the CMA’s taxonomy is, sight should not be lost of the importance of identifying the legal principles to be applied in making those qualitative assessments.
Stricter approach to OCA practices than ‘off-line’ practices?
Many of the OCA practices described in the Discussion Paper apply equally to offline conduct. For instance, a brick-and-mortar store might employ sensory manipulation (using music, lighting, and scents) to induce customers to spend money. Similarly, scarcity tactics can be just as effective in real life (for instance, stores plastered in red posters that proclaim, ‘Limited stock!’ or ‘Going fast!’)
The CMA acknowledges:
‘Competition practitioners have historically tended to regard marketing practices that aim to persuade as pro-competitive, particularly when they allow good products to thrive without foreclosing rivals. But there is also a rich history of issues in traditional marketing practices where businesses cross the line between influencing and misleading, such as making fraudulent claims or exerting undue pressure.’
While this is not a bright line, it nonetheless exists as a legal principle.
However, the CMA seems to suggest that the line should be much stricter in the online world, and potentially that there should be more reliance on ex ante, more absolute rules. Clearly, there are differences between the online and offline worlds which can magnify the harm to consumers and competition, but just how far we should go in shifting the line between ‘influencing’ and ‘misleading’ – and the legal principles to be allied in redefining the line – are big questions to work through.
Who decides – consumers or regulators?
The Discussion Paper opens with the reassuring statement that ‘consumers generally try to make good purchasing decisions, to use sensible strategies to choose between options, and to exit bad deals’.
But towards the end of the Discussion Paper, the CMA suggests that ‘[m]aking consumers aware is..not always sufficient to protect them from harm.’ The CMA says that this is because:
‘There is, however, relatively little evidence that greater transparency through disclosing the presence of choice architecture techniques at the point of final decision making reduces consumers’ propensity to be influenced. In fact, giving more information in the form of disclosures at a later stage only can often backfire, for example, leading to information overload, making consumers unduly more trusting or causing consumers to switch off and avoid engaging in the market altogether. Consumers may also still be influenced, even when they are aware that OCA is being used and are given an opportunity to make a different choice.’
The CMA suggests that the best solution where there is strong evidence of harm may be to ban certain practices outright. In other words, policy makers or regulators will make the decision about what is not allowed. The CMA says it will better equip itself to make judgements about OCA by bringing onboard experts in behavioural sciences and psychology.
Although we agree with the CMA that more research into OCA practices is needed, we should be wary of regulators making qualitative judgments applying vague standards which are not sufficiently moored to legal principles.