Our objective

Understand the central tension between standard economic models and behavioral science

• Principles of economic reasoning (Optimization, Equilibrium) and why this makes persuasion seem hard

• Introduce Behavioral Economics as an approach that accounts for human limitations (limited attention, willpower, etc.)

Political science

A systematic study of political activities and institutions.

• Candidates and voters.
• Many other agents (who may not express their political goals most of the time but could be mobilized during campaigns).
• Groups/coalitions organizing to achieve their goals.
• Behavior can be strategic BUT:
  • We don’t assume all agents are (equally) sophisticated.
  • Agents may be motivated by material pay-offs.
  • And also by non-material (e.g. status) considerations.

“the standard economic view that persuasion is conveyance of information seems to run into a rather basic problem that advertising is typically emotional, associative, and misleading — yet nonetheless effective” (Shleifer 2012)

Rational updating

• Claim or report: e.g. “Inflation rate exceeded 10%”. (R)
• Agents hold an initial view (prior), Pr(High Inflation) = Pr(H).
• Given the report, opinions are updated as follows:

\[ Pr(H|R) = \frac{Pr(R|H)Pr(H)}{Pr(R)} \]

where Pr(R) = Pr(R|H)Pr(H) + Pr(R|Low)Pr(Low)

Pr(R|H) is the essential element.

Is Pr(R|H) accurately perceived?

• You might have heard of confirmation bias.
• Wishful thinking…
• Pr(R|H) could be deliberately ignored if it contradicts cherished beliefs.

More to come about cognitive distortions soon.

Economic Principles

• A large number of “agents” (people)
• Their individual decisions do not affect important systemic parameters
• So prices are “fixed” in the short term
• Markets reach an equilibrium
  • Producers compete
  • Consumers buy what makes them happy

What is economics about?

• Traditionally economics understood as being about satisfying material needs
• No longer accurate today
  
• Achieving goals (broadly defined) efficiently when facing scarcity (or strategic interactions)
• Analysis of constraints, preferences and beliefs to understand incentives

Some problems with the optimization theory

• How long are your time horizons? (Want the dessert now.)
• Expectations about outcomes are unclear
  • Is working out really good for you? Is running as good as yoga?
  • Will I be more productive if I sleep more?
  • Did I choose the field of study that is “the best” for me?

A basic theory makes these predictions

1. Lines will be equally long
2. So in equilibrium, you don’t even need to look. You can assume no line is shorter.
3. If one line looks significantly shorter, there must be a good reason.

What people are minimizing is not some spatial length/position but minutes spent waiting in the line.

Utility is simply the payoff function

U = value of groceries - time costs - monetary costs

• You could still say the model “simplifies too much”
• But that’s the point!
• If needed you can enrich the model (but sacrifice parsimony)

Central Concepts

• Scarcity and Tradeoffs
• Budget constraints
• Opportunity cost (the best alternative use of a resource)

Economic Principles

1. Optimization: people decide what to do by consciously or unconsciously weighing the pros and cons of the different available options.

2. Equilibrium: systems tend toward equilibrium, a state in which no agent would benefit by changing his or her own behavior.

How to deal with uncertainty?

Expectation formation is crucial

Probabilities MAY be objective

• “50% of people like me enjoyed the show”
• Based on the data, the best guess is that my chance of a positive payoff is 50% as well.

Subjective probabilities

• When data is imperfect (almost always)
• How “should” the decision-maker form his or her beliefs?
• How will the decision-maker form his or her beliefs?

Subjective probabilities

• We wish to understand processes to manipulate these objects.
• Major problem: an important component may be unobserved (and non-repeatable).

Example: What is Pr(Social Democrats will increase the minimum wage)?

• To answer, you may need to form beliefs about Pr(SD are committed to raising MW).
• Then Pr(Raise) = Pr(Raise|Comm.)Pr(Comm.) + Pr(Raise|Weak comm.)Pr(Weak comm.)

Economic analysis

• Discover a decision-maker’s objectives
  • How can an objective be achieved with as few resources as possible?
  • (A minimization problem)
  • “Efficient” use of time, money, and other scare resources
• Deviation from optimization is not expected (viewed as non-rational)

Choosing how much to study

Assume that \(U(h) = V - h\).

Choosing how much to study when the outcome is uncertain

Continue to assume that \(U(h) = V - h\).

Then U(0) = - 5; U(10) = -10; and \(U(20) = 0.5 \times 30 + 0.5 \times 0 - 20 = - 5\).

Complications

• Maybe studying (for some classes) “doesn’t hurt”.
• Or some people like to study and others don’t - is it fair to assume everyone has the same preferences?
• Suppose it’s really true that Pr(A) = Pr(B) = 50%. Would it be useful to convince yourself that Pr(A) is higher? (Over-optimism can be motivating…)
• We implicitly assumed that all components (incl. V) are accurately perceived.

Terminology

• Converting hours into grades: that’s the technology.
• How quickly can “hours of study” be turned into good grades? That’s productivity.
• A gain from each hour: marginal product.
• When conversion is faster, it means that studying is “cheaper”
  • For each hour you expend, you get more.
• Optimization just means setting the marginal benefit equal to marginal cost.

Expected utility formula

You can think of any risky choice as making a gamble (used in a value-neutral way here).

In a gamble in which there is a \(p_1\)% chance of winning $X, and a \(p_2\)% chance of winning $Y, the expected take-home payout, or the expected value is:

\[ EV=p_1X + p_2Y \]

Generalizing:

\[ \text{EU(gamble)} = \sum_{i=1}^{N} u(x_i)p_i \]

Choice overload

Setting the stage

• Relatively little work on what the Homer vs. FRP (fully rational person) spectrum means for political competition
  • Voters are inattentive and use shortcuts because the stakes are low (the impact of one vote is microscopic).
  • BUT there are many forms of political participation beyond voting.

Existing research

• Behavioral economics: a domain for various decisions (often personal/financial)
  • Energy consumption and other decisions with environmental externalities
  • Applied and scholarly work about anticipating human biases when designing public policy.
• Cousin fields
  • Marketing: Selling commercial products
  • Psychology: “persuading the self”; staying motivated to achieve your personal objectives

We will view decision science as comprised of these 3 objects

1. Preferences

2. Beliefs

3. Constraints

Changing behavior means that you need to target one of them

Persuasion Models in a Political Context

When there are two prospects, X and Y:

\[ EV=p_1X + p_2Y \]

Conceptualizing persuasive technologies those messages that will change perceptions about any component of the payoff.

How we depart from the standard model

• Beliefs are not necessarily correct
• Drop the assumption that people will act in the same way regardless of how choices are presented
• Do not assume that people care about gains and losses equally

What are correct predicitions?

• Claim or report (R)
• Existing view in favor of a hypothesis: Pr(H).
• Given the report, opinions are updated as follows:

\[ Pr(H|R) = \frac{Pr(R|H)Pr(H)}{Pr(R|H)Pr(H) + Pr(R|\neg H)Pr(\neg H).} \]

• This was a trick, sorry…
• There is not enough information to tell!

Interpreting test results

What we know

• P(+|ND) = 5%
• P(-|D) = 10%
• These are different ways of the test getting it wrong.
• We can calculate complements:
• P(-|ND) = 1 - P(+|ND) = 95%
• P(+|D) = 1 - P(-|D) = 90%

That’s useful, but \(P(+|D)\) is not what we want.

Instead, we want \(P(D|+)\).

Jargon

• P(+|ND) = False positive rate
• P(-|D) = False negative rate
• P(-|ND) = Specificity
• P(+|D) = Sensitivity

Some people will stress “good sensitivity”. That’s not enough…

Interpreting test results

\[ P(D|+) = \frac{P(+|D)P(D)}{P(+)} \]

• There is a way to unpack \(P(+)\), because there are two pathways to observing a positive result.
• But the immediate problem is that P\((D)\), i.e. the prior , was unspecified.

Bayes rule

\[ P(D|+) = \frac{P(+|D)P(D)}{P(+|D)P(D) + P(+|ND)P(ND)} \]

Disease example, ctd.

\[ P(D|+) = \frac{. 9 P(D)}{.9 P(D) + .05 P(ND)} \]

• Suppose that the prevalence of the disease is P(D) = 1%.
• So P(D) is the “base rate”
• If you knew nothing, that should be your estimate that a randomly encountered person carries the disease
• With known base rate:

\[ P(D|+) = \frac{. 9 \times .01}{.9 \times .01 + .05 \times .99} = \frac{.009}{.009 + .0495} = 15.38\% \]

Disease example, ctd.

• Consider a population of 10,000 people
• 100 are sick
• 90 of the sick people test positive
• 9,900 people are healthy
• Among the healthy, \(9,900 \times .05\) = 495 test positive.
• So there are 90 + 495 positive test results.
• Out of the positive results, 90 / 585 = 15.38% are actually sick.

What we just calculated

Base rate fallacy

The correct answer is: \(\frac{90}{(90+495)}\)

Base rate fallacy

Most people underestimate the \(\delta\) term: \(\frac{90}{(90+ \delta)}\)

When they do, the estimated \(P(D|+)^{biased}\) will approach 1, as \(\delta \rightarrow 0\).

Base rate fallacy

• The salient (easily imaginable) cases are on the diagonal
• Key lesson: must remember cases on the off-diagonal

It matters what you condition on

Conditioning on the column

Conditioning on the row

Interpreting the cell entries

Think of these objects as likelihoods of different things you can observe - you only have tests (you never observe the virus)

Think of these objects as hypotheses:

Take-away

You form beliefs about hypotheses based on:

1. signals

2. priors

The likelihood ratio \(LR = \frac{Pr(+|D)}{Pr(+|ND)}\) would ideally be greater than 1.

What we started with

Likelihoods (based on test quality)

Hypotheses

All hypotheses

The correct beliefs, given the test results shown in a given row:

A note on accuracy

So what does accuracy mean?

It’s the ratio of all correct predictions over all classification attempts.

(90 + 9,405 ) / All observations = 94.95%

• The fraction of correct classifications (accuracy) is high.
• But we saw that adjusting beliefs upward too much would have been incorrect.

Useful terms, ctd.

• P(+|ND) = False positive rate
• P(-|D) = False negative rate
• P(+|D) = Sensitivity = True positive rate = Recall
• P(-|ND) = Specificity = True negative rate

False positives / All positive results = False discovery rate

True positives / All positive results = Positive predictive value (PPV), or precision

A popular measure is the F-score: the harmonic mean of precision and recall.

An alternative bias

• Suppose that corrupt politicians “come to mind” easily
• The voter can easily think of examples

Then \[ P(honest | B)^{AVAIL} = \frac{P(h)P(B|h)}{P(h)P(B|h) + \widehat{P(d)}P(B|d)} \]

Given that \(\widehat{P(d)} > P(d)\):

\[ P(honest | B)^{AVAIL} < P(honest | B)^{BAYES} \]

Our workhorse model

• The voter faces a choice before the election
• Let v(.) be the value function.
• The value depends on the election outcome
• v(reelect) = B
• v(change) = 0 or 2 with equal probability
• Expected utility is the same if E[B] = 1.

Some relevant dimensions of the model

• What are voters’ beliefs about B?
• How are those beliefs formed?
• B may have been observed in the past. If B > 1 it suggests that the politician could be competent, but what if she was lucky? Again, there is uncertainty.
• How are probabilities (associated with the uncertainty of supporting someone new and untried) perceived?

What a politician does can be interpreted as a “test”

• A benefit has been realized
• What is P(honest & capable | B) ?
• What is P(corrupt | B)
• And P (corrupt | not B) < 1 (the politician could have been unlucky)

What makes something easily retrievable from memory?

• External reminders (media, politicians…)
• Familiarity
• Importance (a closer friend was injured in a car accident…)
• Personal nature of stories (a friend has a good experience with that brand of the car)
• Recency (if I was just charged a late fee, it’s still salient, but the pain will dissipate over time)

Cognitive bias

A cognitive bias is a systematic error in thinking that occurs when people are processing and interpreting information in the world around them and affects the decisions and judgments that they make.

• Some may be inevitable because of imperfect memory and limits on the mind’s ability to process information
• Inattention
  • Distraction
  • Selective attention

Availability heuristic

What is the probability that, during the next year, your car could be a “total loss” due to an accident?

Availability heuristic

What is the probability that, during the next year, your car could be a “total loss” due to:

• An accident in which the other driver is drunk?
• An accident for which you are responsible?
• An accident occurring while you car is parked on the street?
• One of the above?

When are perceived possibilities more accurate?

• Maybe the probability of a bad accident is overestimated when diverse scenarios are considered
• Then “thinking more” did not improve judgments at all
• A quick and intuitive answer could have been better

Availability heuristic

In 4 pages of a novel, do you expect to find more than 10 words that have the form

• A: 7-letters and n in the sixth position. _____n_
• B: 7-letters and end with ____ing

Most respondents say that B is more likely.

Confirmation bias

a type of cognitive bias that involves favoring information that confirms your previously existing beliefs or biases.

For example, imagine that a person holds a belief that left-handed people are more creative than right-handed people. Whenever this person encounters a person that is both left-handed and creative, they place greater importance on this “evidence” that supports what they already believe.

Myside Bias

A specific type of confirmation bias where people favor information that confirms their own existing beliefs or hypotheses. Involves focusing on the strengths of one’s own arguments while simultaneously concentrating on the weakest aspects of opposing claims.

Projection biases

Curse of knowledge

Once you know something, it becomes obvious to you and you assume others know / should know.

False consensus effect

A tendency to overestimate how much other people agree with us, view the world like us, etc.

Social Projection

Expectations about the behavior of others based on one’s own

. . . Students who cheat on their statistics exams believe that many others cheat as well whereas honest students think that cheating is rare.

Self-serving bias

Its default form

Blaming external forces for bad outcomes, claiming credit when good outcomes materialize

Optimism bias (bad name)

If you want something to happen

Biases, ctd.

Social proof

• Viewing positively what we see others approve of.
• Feeling allowed/entitled to do something if others are doing it.
• Wanting to emulate others

Biased expectations

Poor Affective forecasting

• Typically, people overestimate the affective impact of future events
• Getting used to hardship
• Ceasing to appreciate good outcomes

Anchoring heuristic

Outcome bias

Evaluating a choice based on realized consequences, rather than the quality of the decision at the time

Motivated Reasoning

The tendency to process information in a way that aligns with pre-existing beliefs and desired conclusions.

Rather than assessing evidence objectively, individuals are motivated to arrive at a particular outcome, often by selectively seeking out confirming evidence and dismissing contradictory information.

Outrage Signaling

The public expression of condemnation of a moral transgressor. This can serve to enhance one’s own reputation and signal trustworthiness and adherence to group norms.

Virtue Signaling

The act of publicly expressing opinions or sentiments intended to demonstrate one’s good character or moral standing within a community.

ELM

Key claim: what matters is whether a recipient is motivated to engage with a message

ELM

• Elaboration = engagement in issue-relevant thinking
• The degree to which recipients of a message will reflect [scrutinize arguments, etc.] naturally forms a continuum.

ELM

• Individuals can be persuaded either by the central or the peripheral route.
• If a receiver is prepared to expend cognitive effort then the central (arguments-based) route may produce lasting opinion change.
• Superficial factors, such as likability of the sender of a message play a role when a receiver is engaged via the peripheral route.

Message attributes (other)

• Fear appeals
• Guilt appeals

Considerations: are they ethical? Are they effective?

Message attributes (other)

• Channel factors (face-to-face, vs. digital; audio or video inclusion, etc.).

• Style

• Language

  • Speed, intensity
  • Political-sounding vs. neutral-sounding…

Sender attributes

• Familiarity or reputation.
• Perceived epistemic authority.
• Social endorsements, including engagement metrics online.

Receiver factors

• Prior beliefs.
• Level of involvement with a message.
• Determinants of political knowledge include self-interest and curiosity about politics.
• Contextual factors, such as incentives.
• Demographic characteristics (age, sex, education, digital literacy).

Other theories of communication (related to persuasion)

The “Yale model”

• Asserts that an audience experiences three stages: attention, comprehension, and acceptance.
• As such, “ads work by getting attention and changing attitudes” (Chandler and Munday, 2020).
• But the passive view of the audience is outdated.
• Similarly, the hypodermic needle model assumes that ideas and arguments can be directly transferred from a sender to the intended receiver.

Hypodermic needle model

That’s the idea, born decades ago, that the way to understand the effect of media is to think of it as a direct injection of information, straight into your brain. Your reaction to this information will likely be rapid, predictable, and potent — just as much as a shot of adrenaline to the heart. You learn something terrible about a political candidate; this information injection makes you decide, instantly, not to vote for him.

(Benton, 2023)

Journalists and people around media are sophisticated enough to know that the work we produce doesn’t have huge, life-changing effects on people’s opinions every time we publish.

On the other hand, the hypodermic model can sometimes make media seem all-powerful, which holds a certain allure to those of us who make it.

It can be an especially comfortable mode to fall into when we talk about misinformation and “fake news.”

In general, people think they are sophisticated consumers of information, able to weight new facts appropriately — but other people? Other people? If Facebook tells ’em 2 + 2 is 5, they’ll throw out their calculator.

(Benton, 2023)

Active audience theory

• In contrast to the hypodermic needle model, audiences do not allow senders to impose their meanings on them.
• Recipients are active actors, and their involvement includes interpretations and evaluations of messages (after perception and comprehension).
• This theory rejects the argument that people are “used by the media.”

(Related concept: Oppositional reading.)

Identity-protective cognition

• A variation of directional motivated reasoning. The objective of the agent is to maintain “status in an affinity group”. (Druckman and McGrath, 2019).
• Reactance Theory: When people perceive manipulation or a threat to their freedom, they fight back, arguing against the case they hear (predictions include restorative behavior or aggression).

Discrepancy models of belief change

• People are more likely to absorb a message if their prior views are similar.
• [see also: Confirmation bias; Prior Attitude Effect; Biased assimilation effect; Selective perception].

Perceptual screen

• Acts like a filter that shapes how people process and understand political information.
• Even if two people are exposed to the same data, they may process and remember different information.

Voters need to make inferences about unobserved variables

• Is the incumbent politician honest / corrupt?
• Are the candidate’s policy preferences close to mine?
• If the candidate shares my policy goals, does he or she have have the ability delivery?

Rational formation of predictions would be Bayesian

• Claim or report (R)
• Existing view in favor of a hypothesis: Pr(H), e.g. candidate is of high quality.
• Given the report, opinions are updated as follows:

\[ Pr(H|R) = \frac{Pr(R|H)Pr(H)}{Pr(R|H)Pr(H) + Pr(R|\neg H)Pr(\neg H).} \]

Stereotypes as mental shortcuts

“over-simplified image of a type of person or thing”

(Dictionary definition)

Theoretical debate

View 1: Humans are good intuitive statisticians¹

View 2: Predictions are insensitive to reliable evidence²

One subset of “middle-ground” explanations includes uncertainty-based re-scaling:

• Systematic over-estimation of small values as a rational process (Landy et al. 2018)

What’s going on

• What is P(Red Hair|Irish)? It’s only 10%!

Observations

• \(Pr(Red~hair)\) may be easier to guess than \(Pr(Red~ hair|Irish)\).
• So giving data to people might worsen accuracy of beliefs.
• Ranking or types (hair colors) is the same for both groups.
• But red hair is “easier to recall” for one group, because \(Pr(Dark~ hair|Irish) = 5 \times Pr(Red~ hair|Irish)\).

Representativeness

“an attribute is representative of a class if it is very diagnostic; that is, the relative frequency of this attribute is much higher in that class than in a relevant reference class.” (Kahneman and Tversky 1983)

Distortions in inference

\(Pr(Red~ hair|Irish)\) is mistakenly over-estimated. Why?

Formally, define representativeness as:

\(t_G^* = argmax_{t \in \{dark,light,red\}}\frac{Pr(t|G)}{Pr(t|\neg G)}\)

It follows that the representative color is red: \(t_{IRISH}^* = 10 > \frac{40\%}{14\%} > \frac{50\%}{85\%}\).

Rational updating about group membership

\(Pr(Irish|Red~ hair) = \frac{Pr(RH|I) \times Pr(I)}{\underbrace{Pr(RH|I) \times Pr(I) + Pr(RH|\neg I) \times Pr(\neg I)}_{Pr(\text{Red hair})}}\)

• If P(I) is not properly taken into account, then updating is hindered by base rate neglect..

• Following Bordalo et al. 2016: \(Pr(RH|I)^{st} \neq Pr(RH|I)\).

  • Instead, the true likelihood is influenced by representativeness (and the memory function can take many forms).

BCGS (2016)

\(Pr(RH|I)^{st} = Pr(RH|I) \times \frac{h(R(RH,I))}{\sum{t'}Pr(t'|I) h(R(t',G))}\)

Where:

• R(t,G)=\(\frac{Pr(t|G)}{Pr(t|\neg G)}\) is representativeness of type \(t\) for group \(G\).
• h() is an increasing function

Similar problems

• G = African-American; T = {poor, middle-income, rich}

• G = Democrat; T = {socialist,…}

• G = Republican; T = {alt-right,…}

Stereotypes

If you hear “Democrat”…

… is it easy to imagine a young person/protester/voter?

Are most young voters supporting Democrats?

According to Catalist:

• Among voters age 18-29, the (two-party) Democratic vote share was 62%.

• So yes, most young voters supported Biden.

• What beliefs would a Bayesian hold about the age distribution among Democrats?

  • Need to know the proportion of voters falling into the age group (18-29).
  • Pr(18-29) = 16%.

Rational updating based on a trait

• Conditional on meeting a Democratic voter…
• The probability that they are young is, by Bayes’ rule: Pr(Age 18-29|Voted Dem) = 18.5%.

• But Pr(Age 18-29) was only 16%.
• So a rational agent updated upward, but only slightly.
• A Democratic voter is not very likely to be young.
• The stereotypical association (Dem -> young) would be incorrect.

Is the “typical Democratic voter” young?

• An agent updating according to Bayes’ rule knows the correct answer is “no”.
• But using Bayes’ rule is difficult, even for trained professionals.
• If imagining a young Democratic voter is easy, stereotypes will color the perceptions of demographic compositions.

Which type is representative (easy to recall)?

2020 election data:

\(t^*_{YOUNG}= \frac{Pr(Dem|18-29)}{Pr(Dem|30+)} > \frac{Pr(Rep|18-29)}{Pr(Rep|30+)}\), thus a young voter is representative of Democrats…

Is the typical Republican voter old?

• The answer is again no.
• But by the same logic and calculations, we see that Pr(Older|Rep.) will be overestimated.
• An older voter is representative for G = Republican voters.

\(t^*_{65+}= \frac{Pr(Rep|65+)}{Pr(Rep|64 or less)} > \frac{Pr(Dem|65+)}{Pr(Dem|64 or less)}\)

But a Bayesian would correctly calculate that Pr(65+|Rep) = 26.1%.

In fact, Pr(65+|Rep) < Pr(45-64|Rep) = 35.4%, because more voters are middle-aged than 65+ years old.

The impact on politics?

• Beyond mere misperceptions
• Salience of weights may be manipulated strategically

Factors influencing voting decisions

What is persuasion?

Multiple perspectives

1. Economic
2. Psychological
3. Political

Political messaging

• Should be memorable
• “What are you for?”
• May rely on hints, inside knowledge
  • Reading between the lines
  • So-called dog whistles

Classical economic theory: people hold rational expectations

• People observe data, and update their beliefs (applying Bayes’ rule)

• Persuasion is data presented strategically

• Truth will come to light because the market creates incentives for firms to reveal the truth about competitors: the relevant information will ultimately emerge.

• Problem: Deception is excluded / inadmissible in the formal framework.

• Prediction: some facts will be selectively omitted.

Psychological perspective

• Equivalence-based definition
  • Only if you present the same facts can you engage in re-framing a problem
  • Associated with Tewksbury and Scheufele
• Another school of thought: framing is salience-based - all about selective emphasis
  • Framing as “calling attention to some aspects of reality while obscuring others”
  • Associated with McCombs/UT Austin

Health-related persuasion

• Information alone is often ineffective at changing beliefs and behavior
• Superstition or conventional wisdom can be more persuasive than appeals from authorities
• Evidence from a number of countries is mixed: informing households about contaminated water does change behavior (to some degree), but programs related to sex education, or financial education, do not have an encouraging track record

Kahneman’s Systems

• System 1: your brain’s automatic, intuitive thinking mode
• System 2: careful, deliberate thinking

Faced with bad choices by consumers, such as smoking or undersaving, economists as System 2 thinkers tend to focus on education as a remedy. Show people statistics on deaths from lung cancer, or graphs of consumption drops after retirement, or data on returns on stocks versus bonds, and they will do better. As we have come to realize, such education usually fails. Kahneman’s book explains why: System 2 might not really engage until System 1 processes the message.

– Shleifer (2012)

Motivating vaccinations against COVID

Motivating vaccinations against COVID

Motivating vaccinations against COVID

Motivating vaccinations against COVID

Example of a policy in Slovakia

• A bonus for “recommenders”
• People who register can say who convinced them to do so
• Essentially a commission (broker fee?)
  • 30 euros who people below the age of 50
  • 60 euros if the vaccined person is 50-59
  • 90 euros for people aged 60+

SOCIAL (vs cognitive) persuasion

• Social persuasion

  • Social validation/proof (drink what your friends drink; people go to crowded restaurants, etc.)
  • We imitate people we like (friends/neighbors are best campaigners)

• Authority (church leaders, parents)

• Reciprocity (salespeople offer free samples or small gifts)

• Consistency with the past self

• Scarcity: an opportunity seems more attractive if it is scarce

• Distant persuasion (intentional messaging) can only sometimes be social.

The Elaboration Likelihood Model of Persuasion

Individuals modify their opinions either through cognitive shortcuts (rules of thumb / heuristics / peripheral paths) or effortful reasoning.

Next slide: The Elaboration Likelihood Model of Persuasion (chart from Perloff, R. M. (2017). The dynamics of persuasion)

ELM

Today’s arena for persuasion: Social networking

• We trust recommendations of people like us
• Product recommendations
• Social endorsements (better than explicit product reviews)
• Vendors inform you what people with a similar taste bought (and they test your responsiveness to “special deals”).
• Influencers can persuade others they have (unique) access to relevant information

Modeling persuasion

• Basic idea of a psychologically motivated model of persuasion: sender of message taps into already existing beliefs
• Walter Lippman: persuader tries to create pictures, images, which make people want to do what he wants them to do
• How are images created, and how can we model their effect?
• Key concept: associations

• The persuader needs to tap into the existing map of associations
• Persuasive messages trigger associations by selectively presenting information
• Our minds imagine the rest from memories, stereotypes, past experiences, beliefs