Communicating During The Coronavirus

The “unprecedented new normal”

It would be safe to say you’ve been on a successful digital detox if you haven’t seen at least one of the following in the past week:

  1. The Google trends image about how search trends for the word “unprecedented” have been unprecedented 
  2. An article about the “new normal” and how consumer behavior will change
  3. An easy-to-make sourdough recipe

Unfortunately, we have not yet seen the end of these trends. For many months (or perhaps even a couple of years to come), we will all be discussing this “unprecedented” twist our lives have taken and how we’re dealing with our new normal.

Never before in our lives have organizations and governments had to respond in such short notices to changes taking place around them. And they will have to continue doing so over the next few months by informing us about changes to their services, reminding us about safety measures we should adhere to, and ensuring that the inconveniences we might face are for our own good.

Arguably, none of these are easy to communicate — but understanding how people perceive risks and the psychology behind how people react in these situations can help us create a more effective communication framework.

How we perceive risk

Social scientists and psychologists have been fascinated with risk perception for the longest time. The Extended Parallel Processing Model gives one framework that describes our range of responses to risk.¹ To put it simply, this model explains the process that helps people assess a message about a given risk. When a message is received, people evaluate it at two levels:

  1. The perceived threat of the risk: This is derived from how severe the risk is perceived and how susceptible one feels to the risk.
  2. The perceived efficacy: This has two parts – how one feels about their ability to respond to the threat and how effective that response will be.

As shown in the framework, a low perceived threat would lead to no reaction. A high perceived threat, coupled with a low perceived efficacy, would lead to “fear control,” which often results in the message being ignored. An example of this would be the horrifying stories of people going on spring break holidays, in defiance of recommendations from health officials. The perceived threat of the risk here is high because stories of rising cases are plentiful. The simple idea of staying at home might seem difficult and ineffective for a young person in college living life to the fullest around friends. This idea leads to a “defensive motivation” approach —“I would rather just ignore this threat and enjoy my spring break.”

Source: Adapted from Putting the fear back into fear appeals: The extended parallel process model²

A high perceived threat, with a high perceived efficacy, would lead to “protection”, where the individual does everything in their power to control the situation by following precautions. For instance, wearing a mask or maintaining physical distance. In this case, the threat appraisal is high, but people believe in their ability to control the risk with small steps such as wearing a mask, and hence, put themselves in the “protective motivation” approach.

While this is easier said than done, communications do play an essential role in this assessment. Whether someone reacts with fear or anxiety (or does not react at all) could be driven by how risk and coping mechanisms are communicated to them.

Framework for Communication:

Based on behavioral science concepts, here are some ways to communicate better during the pandemic. This is not an exhaustive list, but a good starting point to analyze communications.

1. Transparency: Some of the most intense discussions around COVID-19 have been around false information. One golden rule of communication is that if you don’t communicate, people will assume. And when people assume—especially when they are in fear—they will only assume the worst. It is thus essential that businesses and governments are open and transparent about facts. For example, the Government of Singapore has received praise from the WHO² for its transparent information sharing through Whatsapp and Telegram.³ By doing so early in the crisis, the government built trust, which allowed them to communicate sensitive messages more effectively. 

Presenting the right information in the right way also lets the receiver perceive the risk more appropriately. For instance, using the flu as a reference turned out to be a poor way of getting people to appraise the risk of COVID-19 because the flu is a known and smaller risk.

In the next few months, people may need convincing when it is time to return to normal; and in this time, information and transparency can play a critical role. If Chinese food delivery apps are anything to go by, this indeed is true. From showing the temperature of the delivery person on the app to giving people factual information when a case surfaces in the service industry, businesses must be prepared to be truthful.

Chinese app showing the temperature of driver
Source: Twitter⁴
Indian delivery app Zomato coming forward with news of delivery person being infected
Source: Twitter⁵

2. Shared Values: It’s an understatement to say that our generation has never before faced an existential threat like COVID-19. And threats like these that are as impactful to all of humanity can lead people to have feelings of collective angst.⁶ This can result in good, such as when people have feelings of selflessness in which they wish to do everything they can to help society. It can also lead to negative consequences, as we saw with the case of panic buying.⁶ To channel this emotion the right way, businesses and governments need to bring up shared identity, so that the risk of an existential threat is seen as something people can cope with.⁷

Shared identity here refers to a sense of connection people feel to each other through shared experiences, interests, or even common sufferings. While the shared experience of being in a pandemic may be strong enough to evoke the feeling of being connected as many of us move on without getting infected, this shared value may be insufficient alone. Getting people to act in the interest of the whole group’s safety might prove to be even more challenging in such a scenario. Communication will then play a crucial role by reminding people of a shared identity when it does not exist.


The AI Governance Challenge

The Budweiser One Team campaign is an excellent example of doing just that. In light of the COVID19 crisis, the parent company of Budweiser, Anheuser-Busch InBev, redirected $5 million of its sports and entertainment marketing spend to the American Red Cross to support the fight against the pandemic.⁸ The One Team campaign shared this information with viewers through a compelling video. While there have been criticisms about brands jumping on the COVID-19 bandwagon, Budweiser jumped on it with great insight—by understanding that their consumers want to belong to and support sports teams. By showing sports teams composed of everyday COVID-19 heroes, they built a new shared value. 

Source: Budweiser⁹
Full ad:

Another example of using a shared value to channel emotion the right way is the case study of hand-washing communications shared by the Behavioral Insights Team, a social purpose organization that applies behavioral science to public policy.¹⁰ The team tested seven hand-washing communication posters (that each originated from different countries around the world), with 2,629 respondents in the UK. The most effective poster turned out to be one made by Taiwan. Take a look and see how impactful the use of “you and me” is in the first line of the poster.

Taiwan CDC Handwashing campaign
Source: BIT¹⁰

3. Social Norms: One aspect of communication that has worked well for a long time is the use of social norms in nudging behavior. Now more than ever, we need to remind people about the norms they should follow. Once the lockdowns end and people try to get back to their lives, the pandemic might not be as great of a focus as it has been thus far; and yet, we still will need people to adhere to safety regulations. This is important for the risk coping appraisal to be a success and for people to feel confident in their ability to keep themselves safe. One way to do this is to make the new norms more salient. Celebrities and leaders adopting new techniques for greeting each other—such as using elbow bumps to say hello—is an excellent example of this.¹¹ Communicators will need to assess how to make new norms common, accessible, popular, and easy to follow. Behavioral design lessons might come in handy here, especially if adopted to the cultural contexts where they are implemented.

Physical distancing with graffiti at a market in Chennai, India
Source: qz.com¹²

While this list of ways to communicate is in no way exhaustive, it is a starting point for organizations and governments to start thinking about how to address their audiences in our new normal. As we brace ourselves for hard times, we must remember that communication is one thing that will always remain with us. With science-backed insights about how people react to risk and what we can do to help them, we might just be able to convert communication into a strength that no virus can touch.

How Facebook Increased The Number Of Votes In The 2010 US Congressional Elections

Imagine you open your Facebook page and see a fundraising program for koalas burnt in the Australian bushfires. You see that Sarah (your best friend), Tom (your neighbor), and John (your friend’s colleague) have already donated to the cause. You click the link to donate and announce your participation by pressing the button so that everyone knows you support the same cause. Had you financially planned for that donation in advance? Probably not, but you like Sarah, so if she did it, maybe you should too.


This hallmark study,¹ which sought to investigate if social media can increase voter turnout, was performed in the US in 2010 and published in 2012 in the prestigious journal Nature. It has since been mentioned more than 50 times in different blogs and news sites, including The Guardian. Despite the widely-held belief that online mobilization played a significant role in recent elections,² the results of a meta-analysis study on email experiments indicate that online methods are ineffective for influencing voter behavior.³ What makes social media so attractive to researchers is the possibility of reaching large populations that it brings; even an increase in the number of voters as small as 1% can be substantial given a large enough population of users.


To verify the hypothesis that voter turnout can be increased with the help of an online social network, the researchers randomly chose about 61 million eligible voters who had access to Facebook on November 2nd, 2010, the day of US congressional elections. Individuals were randomly assigned to either a control group, a social message group, or an informational group, and the actual voting rate was collected for the individuals who participated in this study using public voting records.


The social message group, in which individuals saw their friends’ faces randomly, pressed the “I Voted” button about 3% more than the information group. This indicates that merely seeing familiar faces can impact an individual’s voting behavior. This group also clicked on the link that provided information on where to vote 0.26% more than the information group, suggesting that seeing familiar faces has a small yet positive impact on information-seeking behaviors.

In terms of real votes, the social message group voted 0.39% more than the control group, suggesting that seeing the faces of friends can contribute to an increase in real-world voting. Interestingly, the number of real-world votes for the social message group was the same as that of the information group, suggesting that seeing pictures of friends isn’t enough to increase real-world voting in and of itself, even though this increased the number of people who said they voted.

The indirect effect of “Facebook treatment” on users’ friends

To better identify the contagion (i.e., the indirect effects that spread from a person to another because of being friends), the researchers used the number of Facebook interactions between each pair of friends as a proxy for their closeness. Therefore, the researchers could test whether the closeness of friendships can change the voting behavior of individuals. The results showed that the probability of voting increases in individuals whose close friends had voted.

Close friends affect an individual’s voting behavior. As the level of interaction increases, so does the effect of a friend’s voting behavior on an individual’s voting.

The results of this study demonstrate that while friendships increased political self-expression (i.e., the act of saying that you’re voting), close friendships—although they make up only 7% of Facebook friendships—accounted for a significant increase in real voting. It has been estimated that Facebook social messages increased the number of votes directly by about 60,000 votes and indirectly (i.e., through social contagion) by another 280,000 votes.


The AI Governance Challenge


The results of this study are apparent; however, there have been concerns about potential privacy violations and the ethics behind web-based experiments.⁴ This study showed that online political mobilization increases political self-expression as well as real voter turnout. But these results go beyond this context and point to the importance of social approval on behavioral change. It should be noted that the indirect effect of the message was about four times more potent than the direct effects, highlighting the influence of social networks in society and their ability to change individuals’ behavior.

Harnessing the social approval effect in real-life

Policymakers could leverage the social approval effect to drive change in many different areas, including:

  1. Social media. Social media may provide political decision-makers with a cheap, large-scale method of impacting individuals’ behavior both directly and indirectly. As an example, sending a message on social media about receiving a vaccination and referencing users’ friends who have gotten the vaccine (i.e., showing their profile pictures) could increase the vaccination rate.
  2. Public health. The social approval effect could be leveraged by encouraging individuals who are willing to wear a mask when in public places to post a photo on their social media accounts. This, in turn, could encourage their direct and indirect friends to wear masks as well.
  3. Energy conservation. The social approval effect could encourage a reduction in energy waste. Simply knowing your neighbor’s level of energy consumption could affect your behavior enough to reduce your consumption, shown by the pilot study that OPOWER, a US-based software company, performed in partnership with the utility providers of several American states.⁵


The efficiency of the social approval effect can be explained by the concept of conditional cooperation, which means that people are more prone to participate in an act of public good — such as receiving a vaccination, wearing masks in public places or consuming less energy — if they see that other people are doing the same. Merely seeing that others are acting in a certain way can indeed have a profound impact on our choices.

Harnessing Social Norms for Social Good

When I wrote this article, the Coronavirus had just been declared a pandemic by the World Health Organization,¹ and the United States had declared a national emergency.²

Since then, health officials have tried (quite successfully) to promote behaviors that can reduce the spread and transmission of the Coronavirus — including the one we are all quite familiar with now, social distancing, or the practice of avoiding unnecessary contact with others in order to slow the transmission of the disease.

While social distancing is essential to slow the spread of COVID-19 — which in turn reduces the strain on our healthcare system and saves lives — it is personally costly for the individual, as it requires personal sacrifice (forget about going out to brunch).

As such, now more than ever it is important to understand how insights from behavioral science can promote prosocial behaviors and get people to prioritize shared over individual interests. One potentially powerful set of tools to promote behavioral change is social norms.

Social norms are overt or unspoken rules that govern what behaviors are viewed as appropriate in society. Norms are frequently enforced by other members of the group and violating a norm can result in anything from social disapproval or informal sanctions to ostracization.³

Our predilection to create and conform to social norms is thought to be universal, occurring in diverse societies and cultures across the planet, although their expression is likely culturally dependent.⁴ Children as young as 3 years old begin to shift their behavior in accordance with local social norms, suggesting that our sensitivity to norms arises early in development.⁵ Taken together, this early internalization and ubiquity of norms suggests that we have evolved an innate sensitivity to follow them.⁶

Research on social norms has found that they have a profound impact on behavior across a wide range of domains such as decreasing littering,⁷ improving energy conservation,⁸ increasing hygienic behavior,⁹ and reducing student gambling.¹⁰

To get a better sense of the influence of social norms on human behavior, let’s look at a foundational study of social norms by Robert Cialdini and colleagues.¹¹ 

An Illustrative Example

Petrified Forest National Park in northeastern Arizona is a vast swath of desert and colorful badlands known for its abundance of 225 million-year-old fossilized trees. Unfortunately, however, the park faces rampant theft of its’ petrified wood from visitors, which threatens to compromise its natural beauty.

In a bid to counteract this theft, the park teamed up with Cialdini and colleagues to see if they could help simply by rewording the messaging of signs around the park. 

The authors manipulated whether signs displayed an injunctive norm, which refers to what is commonly approved or disapproved of (“please don’t remove the petrified wood”), or a descriptive norm, which refers to what is commonly done (“many past visitors have removed the petrified wood”). They also manipulated whether the norm was strong, such that it was negatively framed (“please don’t remove the petrified wood”), or whether the norm was weak, in that it was framed positively (“please leave the petrified wood”).

They found that people were significantly less likely to steal petrified wood from the park after seeing a sign that displayed an injunctive norm compared to a descriptive norm, with the lowest level of theft occurring when the sign reflected a strong injunctive norm.

The results of this study suggest that messages that convey a descriptive norm about what is commonly done (e.g., “many visitors have removed the petrified wood”) can backfire, resulting in greater levels of that behavior. This is because descriptive norms convey the message: “if everyone else is doing it, then it’s probably okay if I do it too”. Reframing this message around the injunctive norm, what others approve or disapprove of, is much more effective at changing behavior.

Practical Implications

While this research is hardly hot off the press (the study described above was published in 1990), organizations and health officials have been slow to apply its insights. This is surprising, given that simply reframing messages in terms of social norms can offer a cheap but effective behavioral intervention.

Exactly how can we harness social norms for societal and organizational good? Research suggests three main ways to effectively utilize social norms.

First, as seen above, messages framed as injunctive norms have the strongest impact on behavior, above and beyond descriptive norms. For example, if you wanted to promote energy conservation by having hotel guests turn off their room lights, you should emphasize that most people disapprove of leaving lights on and wasting energy (an injunctive norm), rather than emphasizing that most people leave their lights on and waste energy (a descriptive norm).

Second, messages have a stronger impact on behavior when they are framed negatively, describing which behaviors are not approved of. In keeping with the energy conservation example, it is better to have a message that says: “save energy, do not leave your lights on”, rather than one that says: “save energy, remember to turn your lights off”.  

Lastly, while you can make norms explicit through a sign or message, often norms are not so clear and salient, making them less effective at changing behavior. Norms are most effective at changing behavior when they are made focal in attention and brought to our conscious awareness — such as when explicitly stated on a sign or clearly modeled by another person. For example, if you see another hotel patron model the norm by turning their lights off as they leave their room, this makes the norm salient and would likely make you more prone to shut your lights off too.  

Using Norms to Promote Social Distancing

With all that in mind, let’s return to the question at hand: how we can use social norms to promote social distancing?

First, we should describe messages urging social distancing as an injunctive norm by framing them in terms of what others approve of (e.g., “most people think social distancing is the right thing to do”). Indeed, framing messages about social distancing as a descriptive norm (“most people are not social distancing, contributing to the spread of the disease”) might actually be detrimental.


The AI Governance Challenge

Importantly, if conveying both descriptive and injunctive norms, the norms should align, as it would be counter-productive to have the normative information conflict.¹¹ For example, communicating that most people aren’t social distancing but approve of others who are would likely result in less social distancing relative to messages that convey consistent normative information (i.e., “everyone is social distancing and approve of others who are too”) or only conveys the injunctive norm.

Second, we should frame the injunctive norm negatively (“do not go to gatherings of more than 25 people”) rather than positively (“only go to gatherings of less than 25 people”). This distinction is subtle yet powerful, and can result in significant differences in behavior. 

Lastly, it’s important to induce a normative focus so people are more likely to remember and follow the norm. That can take the form of a steady stream of reminders and messages from the media and public health experts. Influencers and celebrities could also model the norm of social distancing via social media, such as this video currently trending on Twitter featuring Mel Brooks.

In sum, research on social norms offers powerful insights for organizations and public health officials on how to promote prosocial behavior. In a time of global crisis, these insights are more important than ever.

Evidence-Based Strategies For Washing Your Hands

Have you washed your hands today? Beyond the widespread practice of social distancing, health officials recommend frequent hand washing, cleaning high-touch surfaces, and no face touching to help limit the spread of COVID-19. In light of these recommendations, behavioral scientists have compiled evidence-based strategies to increase hand hygiene. Complementary to these efforts, the EAST framework (Easy, Attractive, Social, Timely) can help practitioners adapt these strategies to their own contexts.

Make it Easy. Convenience and simple messaging are critically important to successful public health campaigns and can reduce the amount of effort required to follow hand hygiene recommendations. Hand sanitizer use increases substantially when freestanding dispensers are placed near hospital entrances and are easily visible (Aarestrup & Moesgaard, 2016; Cure & Van Enk, 2015), and automatic paper towel dispensers boost hand washing frequency (Ford et al., 2013). Simple posters with bright colors, step-by-step visual instructions, and minimal text are most effective to enhance knowledge of proper hand washing procedures.

Make it Attractive. Scents, colorful visuals, and rewards can unconsciously guide users to wash or sanitize their hands. In a hospital setting, a citrus smell more than tripled hand sanitizer use (King et al., 2016). In schools and universities, colorful stickers of arrows and footprints placed on the floors lead to more frequent hand washing (Blackwell et al., 2017; Dreibelbis et al., 2016). At home, children are four times more likely to wash their hands when toys are visibly embedded inside their soap (Watson et al., 2019).


The AI Governance Challenge

Make it Social. Hand washing more than doubles when other people are watching. Normative messages and cropped photographs of human eyes increase hand washing and sanitizing behavior (Aarestrup & Moesgaard, 2016; Judah et al., 2009; King et al., 2016).

Make it Timely. It’s important to remind users at the right time. In the examples mentioned above, arrows and footsteps guided adults and children to the sinks immediately after using the toilet, and hospital visitors were reminded to sanitize their hands as soon as they entered the hospital and before visiting a patient.

Finally, using a combination of strategies instead of a single strategy has proven to be more effective at increasing hand hygiene (Gould et al., 2017; Huis et al., 2012; World Health Organization, 2009). For example, placing sanitizer dispensers near hospital entrances (Easy) increased compliance from 3% at baseline to 20%, but combining the placement strategy with a bright red sign (Attract) and a normative message (Social) increased compliance to 67% (Aarestrup & Moesgaard, 2016). As we start transitioning out of formal lockdowns, it is important to think about how you can promote hand hygiene behaviors in your home or your organization. Following these simple steps is a good start.

Does Anchoring Work In The Courtroom?

Anchoring is one of the most prevalent and enduring heuristics that decision-makers encounter in their daily lives and is particularly powerful when making decisions under uncertainty. An anchor is an initial reference point that has an outsized impact on how decision-makers interpret and encode subsequent information on the topic. And most strikingly, we struggle to overcome the anchor’s effect even when given incentives to do so or when made conscious of the resulting bias.

Anchoring: A Simple Example

Take for example a study where participants were asked to recall the last two digits of their social security number and then were subsequently asked to price a bottle of wine. You can already guess the outcome: those with social security numbers ending with high digits (think 70s, 80s, or 90s) were willing to pay more for the wine than those with social security numbers ending with lower digits.

[Image taken from]

But does anchoring work in the courtroom too?

Anchoring Effect & Juries

Juries are just regular people without any specialized legal expertise, so you’d expect that just as regular people are influenced by anchors in pricing wine, they might also be influenced by anchors in the courtroom.

And you would be right. 56 mock jurors were presented with a hypothetical case where the plaintiff was arguing that her birth control pill caused her ovarian cancer. She was suing the Health Maintenance Organization (HMO) for prescribing her the pill. In the high anchor group, the plaintiff asked for $5 million in damages; in the low anchor group, she asked for only $20,000. The question was: is this anchor going to affect the jurors’ perception of causation?

The answer, predictably, was yes. Jurors in the low anchor condition were 26.4% confident that that the HMO caused the injury, whereas jurors in the high anchor condition were 43.9% confident that HMO caused the plaintiff’s injury.

[image from]

Anchoring Effect & Judges

Judges, unlike juries, are subject-matter experts. Can they really be susceptible to anchoring as well?

According to some preliminary research, the answer is probably, yes. Once an anchor is set, research suggests that a judge is more likely to interpret subsequent information around that anchor, even if the anchor is totally irrelevant.

In one study, judges were presented with a hypothetical case involving a shoplifter who had just been caught for the 12th time. The judges were asked to sentence the shoplifter, but only after the prosecutor made a sentencing demand. And here’s the twist, the judges were told ahead of time that the prosecutor’s demand was totally arbitrary and random; therefore the prosecutor’s sentencing demand contained no useful information.

Even so, the judges who received the low anchor (i.e., the prosecutor demanding a shorter sentence) landed on a shorter average sentence than the judges in the high anchor condition.

[image from]

Ok, so the judges assigned weight to a prosecutor’s (random) sentencing demand. But does this really prove the anchoring effect in judges?


The AI Governance Challenge

So, to squash any doubts about the anchoring effect, the same group of researchers went about designing the most absurd scenario possible. Here, a group of judges were given a hypothetical case where prosecutors were charging a defendant with theft. Instead of being given the prosecutor’s sentencing demand, the judges were told to discover the sentencing demand by rolling a dice. (Yes, really.) The dice was rigged so as to land on high numbers for one group of judges and low numbers for the rest.

And … somewhat unbelievably, the outcome of the dice-rolling exercise influenced the judges’ sentencing decisions.

[image from]

Ok, so even if the results of this study hold, judges don’t roll a dice before they make sentencing decisions. And prosecutors don’t make random sentencing requests. So does anchoring really affect judges’ decision-making on the bench?

Well, it might. Englich et al. explain:

“Even though judges typically do not throw dice before making sentencing decisions, they are still constantly exposed to potential sentences and anchors during sentencing decisions. The mass media, visitors to the court hearings, the private opinion of the judge’s partner, family, or neighbors are all possible sources of sentencing demands that should not influence a given sentencing decision.”

[Playing Dice With Criminal Sentences: The Influence of Irrelevant Anchors on Experts’ Judicial Decision Making (2006) by Birte Englich, Thomas Mussweiler, & Fritz Strack]