Notre Dame part 4

Ann Schlosser (University of Washington, Michael G. Foster School of Business), What are my Chances? The Persuasive Implications of Using Imagery to Interpret Ratios

Commentators:            Raghu Rao (McCombs School of Business, University of Texas at Austin)

How to present ratios to consumers, perhaps in public service announcements about certain risks.  Turns out that numbers matter: 1 in 10 is perceived differently than 10 in 100.  Rao’s background is game theory, so this was all new/surprising.

One explanation for why: denominator effect, where the denominator is hard to process—total of affected and unaffected people, whereas the numerator is easier to process. So people will focus on the numerator.  100 in 1000 will convey higher risk. Another explanation: people start with the numerator, so the first thing you encounter has an anchoring effect. 

Question: should we talk about reductions in risk in the same way?  So if we’re trying to convince people to take acts that will minimize their risks, how do we do that?

Another question: what is the boundary condition here—10,000 to 100,000—if more is more effective, should we keep increasing the numbers?

Prospect theory makes some claims about how we look at probabilities. Robust finding: People often overweight the low probabilities and underweight the high probabilities.  If we look at that and the ratio, is there a connection between those two findings?

Rebecca Tushnet (Georgetown)

Interactions between numbers used in text, imagery, and the emotional appeal of an ad—whether it made the viewer feel helpless or efficacious.  All of these turn out to affect beliefs and intentions.

Questions about the study itself: I wanted to know more about how the studies assessed whether participants were using imagery.  According to the paper, participants reported their agreement with four statements assessing imagery processing, e.g., “I processed the ad using imagery,” but I don’t know whether those are good measures—do respondents really know what experts mean by “I processed the ad using imagery”?  Even if they do, do we believe them when they say how they processed the ad?  This is one area where asking the questions may influence the reporting.

Talk more about how specific emotions were evoked and measured. The studies suggested that “challenge” was an important emotional reaction, compared to fear, hope, guilt, or regret.  How did the ads trigger these emotions?  Were there particular groups of subjects who reacted as expected and others who didn’t?

More general observations: There is a divergence of questions asked by marketers and marketing researchers and those asked by lawyers.  Risk of misunderstanding (which I’ve documented with respect to studies about so-called trademark dilution).  When translating, can be important to define imagery carefully: it can include pictures in the ad itself, but can also include processing in which participants form mental images.  To a lawyer, it wasn’t clear throughout the paper that imagery meant both pictures of cars and also low ratios, which apparently caused consumers to imagine people represented by the numbers.  Since ads almost universally use pictures, if we’re talking about marketing regulation, then we may not need to specify what we mean by imagery, but I suspect it would still be a good idea.

Another concern, certainly known to marketers: intentional manipulation.  Lauren Willis has investigated how banks have responded to consumer-protective changes by manipulating decision environments to get them to “opt in” to overdraft protection.  How will understanding of the effects of different presentations on risk perceptions be used?  Will food companies use it to convince us that our risks of harm from obesity are low by carefully using smaller numbers when those will be more lulling? 

Regulatory implications: if presentation matters so very much, this suggests that regulators must be very careful about the level of generality of regulations requiring disclosures or other information.  The two alternatives would appear to be command and control—no, you don’t get to choose whether you use small numbers or big ones in your ratios—or outcome-based measures: you can structure the disclosures as you wish, but they must pass a copy test indicating that some significant percentage of consumers received the intended message.  Both of these have their weaknesses, but given the manipulability of information and the strong incentive of regulated parties to evade uncomfortable regulations, one or the other must be considered.

Finally, inseparability of emotion from knowledge.  The emotional effects of the ads changed how consumers thought, sometimes in the direction of a better understanding of the facts.  This is an important point for marketers to continue to convey to lawyers, judges, and policymakers, who tend to take a ridiculously rationalist viewpoint that assumes that emotion can be separated from reason.  Latterly this idea has been used to strike down cigarette warnings because they used imagery and were supposed to generate negative emotions.  But disease and death, the predictable consequences of smoking, should generate negative emotions; information without emotional tone will not inform decisions.  (See also: black box warnings on drugs.) If our commercial speech doctrine doesn’t soon start recognizing this, the regulatory state will be in trouble.

Schlosser: Reduction of risk would be a great extension.  Would think that large numbers would still lead people to see a dramatic drop, but “from 15 to 5” is an interesting thing to test.  She looked more at the lower bound—a small group is easy to spontaneously imagine.  If you went more extreme, there might be a point where respondents might not understand what millions and billions are in relation to each other.

Prospect theory: typically research on ratio has focused on low probability events, because with high probability events the effect seems to wash away and people recognize the high probability. Also it’s really hard to communicate a 10% risk.  Can tiny risks be presented in ways that produce meaningful differences? The evidence suggests they can.

Rao: imagery seems to flip results in certain cases—what’s going on?

Schlosser: holistic processing—may shift loss v. gain frames.  People think less about the probability of being in a car accident and then more about the probability of not being in a car accident. 

McKenna: also a question of which side of the ratio you’re presenting. If you present “4 out of 5 people have no problem” do you see different dynamics?  Which gets to the regulatory point.

Heymann: risk aversion also comes in.

Schlosser: two different processing mode, intuitive and analytical.  These are simultaneous.  People will realize that one is mathematically right, but they go with the one that feels right.

Montgomery: Imagery and discursive processing—you show a preference for a smaller denominator. On board w/the argument that this is because it’s easier for people to visualize. If you make it easier to visualize the denominator, would that go away? If you make it more difficult to visualize 4 people, can you flip it?

Schlosser: 1 in 4 may be difficult to avoid visualization.  Noticing when you intended to prompt imagery—one thing that reduces imagery is to have people think analytically and logically, which might depress the effects. 

Montgomery: think of 4 people you know—change the emotional valence.

Goldman: regulators in this space proceed with zero knowledge of social science; 1% knowledge may be worse.  How can you use this research for evil? You can imagine comparative advertising where one presents it as 6 out of 10 and the other says, 40 out of 100.  This paper explains why consumers take away different messages despite literal truth—it’s simple manipulation of consumer processing.

Schlosser: thought of this as a way for nonprofits to convey good messages.  The policy choices may be to tell people how they have to present numerators v. denominators. Or to require images.

McKenna: is there literature about processing ratios instead of percent?  Does it matter if you represent 1 out of 3 as 33%?  Measuring expected behavior: it’s hard to measure expected purchasing behavior because what people say they’ll do and what they will do is different.

Regulatory: if you require everyone to have the same baseline. Study shows that as an absolute matter, fewer consumers understand information conveyed in one way v. another. Is there any indication that consumers would learn better over time?

Schlosser: ratios are typically much more effective at conveying risk; prior research establishes that percentages are too abstract.

Rao: especially when percentages are low, ratios are likely to do better.

Schlosser: some offer an evolutionary explanation for this: we are used to processing frequency and not percentage.

Expected behavior v. intentions: there are some creative ways to get at that.  How can I track people? IRB difficulties (false memories are ok, even though they won’t be dispelled by disclosure, but apparently not sunscreen!).

Rao: One large-scale study tracked people; very controversial showing that ads improved intention but had no effect on action; industry argued that there were methodological problems.

Schlosser: very hard to push ingrained behavior.  Even a little nudge towards “this is a problem” is a move in the right direction.

Sheff: Failure of mandatory disclosure is well-known. There may be ways to avoid some of the standard failures with proper design.  CFPB is working on better disclosures too. Disclosures of side effects/efficacy in medical settings. Formalized framework sensitive to features of cognition.

Yen: Curious whether research could be extended to product design where victim precaution matters.  Failure to warn is available if presentation of warning wasn’t reasonable. Reasonableness on producer side might require some account of how consumers react to information: bland information panel might not be enough. Consumer precaution: we may also say it’s consumer’s job to push through the ratios and understand that 3 in 10 and 30 in 100, but that would be a big change in the law.

Is it possible to research verbal v. numerical cues? This is very dangerous v. 90% chance of injury.

Also: how is this related to brand commitment? If people are committed to their brands, will they react to numbers at all?

Schlosser: suspects that people will switch numbers in memory in ways that are congruent. Will they counterargue/switch over to giving different numbers (e.g., ok 4, out of 10 people who regularly eat at McDonald’s get sick, but 6 don’t).

RT: on expected behavior: I immediately thought about the Obama campaign’s attempt to convert intention to action, not for nothing taken from marketers like Cialdini.  “Make a plan” as effective predictor of action.  If they indeed ended up imagining themselves putting on sunscreen, for example, more of them probably did so, but that’s imagery-mediated.

Schlosser: challenge as the mediator suggests a measure of personal responsibility—it looks like it’s imagery processing that’s really doing this. Ease of simulating the behavior mentally is key.

RT: so do ratios matter at all?

Schlosser: interesting extension to see whether people who processed it textually behaved differently than those engaged in imagery.

Montgomery: sufficiently painful imagery may backfire because people don’t want to confront it.

McKenna: with smoking images, too.

Schlosser: in some ways, numbers can also validate claims, so “very dangerous” may have trouble being convincing.

Montgomery: even numbers are relative in a way.  People judge based on their own situation—easier to visualize out of 4 than out of 10.

Schlosser: but that’s with smaller numbers. When the numbers are larger, people are overinfluenced by which number comes first (denominator neglect).

RT: I am starting to wonder if we’re just rearranging deck chairs on the Titanic: if disclosure doesn’t do much, we might need bigger design changes—narrowing the streets instead of making the speed limit sign stand out more.

General discussion of anchoring effects and their implications for statutory damages—even when a judge thinks s/he’s giving a defendant a good deal, large statutory damages may do harm.

Manta: similarly, prosecution adding charges to multiply exposure.