In our previous behavioural science blog post, we covered the last bit of theoretical ground on system 2 errors. We discussed cognitive load and how to manage it within your choice architecture, for an all around better customer experience. We also ran through the theory behind confirmation bias and human tendencies towards bad mental accounting. If you haven’t had a chance yet, you can read the article here.
This time around we’re changing step. So far, we’ve covered a lot of theory, with some practical insights dotted throughout. You should be familiar with the basics of dual process theory, heuristic bias and spotting different kinds of errors. Hopefully you’re already beginning to employ those insights within your company! In this post, we’re going to focus in on the practical insights we can garner from two specific phenomena: semantic prosody and mindless choosing.
By the end of this article, you’ll have a better understanding of the importance of linguistics, particularly within a global business or international customer journey, as well as the importance of providing sufficient reasons for calls to action.
Let’s start with semantic prosody…
Semantic prosody:
I’m going to ask you to read a couple of sentences. I want you to think carefully about how they make you feel. Before you do, imagine you’re hearing these statements from a doctor. The first sentence is as follow:
“Surprisingly, ingestion of the substance causes endocrination of abdominal lipid tissue.
Think about your gut feeling and make a mental note. Now the next one:
“Surprisingly, ingestion of the substance produces endocrination of abdominal lipid tissue.”
How do these sentences make you feel? What does each sentence mean? Do they mean the different or the same things? Ask yourself these questions and search for the difference between the two statements.
Unexpected Feelings?
Overall, people seem to overwhelmingly think that the first option is worse to hear from a doctor than the second. The study these excerpts are taken from, shows a huge gap in good versus bad evaluations of these statements! Across multiple samples of around 300 people, between 20-35% view sentence 1 as more negative than sentence 2, with results consistent across 3+ repeat measures studies[1].
Biologists and anatomists will know that there is no such phenomenon of endocrination of abdominal lipid tissue, and that actually these sentences make no sense at all.. But a sentiment is delivered and the words delivering the differing sentiments here are ‘cause’ and ‘produce’. The difference between each coming from a phenomenon that linguists call ‘semantic prosody’, whereby people develop negative and positive associations to words, based on the context within which they most commonly hear them. We may consider this association a kind of type 1 error and a form of heuristic bias.
These statements are designed by the experimenters to elicit different emotions, through the use of semantically prosodic language. Cause is the classic example of a prosodic word: “Smoking causes cancer”. “The cause of death”. Cause is a neutral word which we most often hear in a negative context. As a result of heuristic bias and semantic prosody, this can lead to certain phrases and sentences as being evaluated as negative, when in reality the intended meaning was neutral.
Produce, on the other hand, is more often associated with positive activities. Hauser et al believe there are many more examples of prosodic language, but as of yet they’re still understudied.
Implications
So, what’s the takeaway from semantic prosody? Whilst this is culturally specific language, and there’s no exhaustive list of prosodic words, Hauser et al’s study raises an important issue when developing influential conversations with your users or customers. Choose your words carefully! The right collection of prosodic words can sway a customer in a positive direction, but if poorly designed, the opposite is also true.
At ContactEngine, when we design conversations, we take care to share our messaging internally to avoid situations like this. We perform robust sanity checks, to help better understand the context of the messages we want to send and to make sure these messages align with the context of the broader conversation. We also perform systematic a/b testing, to make sure we are always using the most effective linguistics. Sometimes that means utilising semantically prosodic words, and other times it means avoiding them.
Working across the UK, EMEA and North America, we have to bear in mind the different cultural context of words in different places. Language in text format is inherently ambiguous, and so we must be aware of geographic nuance. A poorly understood, semantically prosodic message, could be the difference between an engaged or a turned off customer. When operating at scale, you can’t afford to look past these kinds of biases.
The photo copier study
Understanding the content and context of the language you use is important. But the structure of that language is equally important. In perhaps the best example of its kind, Langer, Blank and Chanowitz[2] demonstrated the importance of offering reasons within calls to action, and the effect this has in producing ‘mindlessness’ – which means something like insensitive compliance to calls to action. Their findings are fascinating.
First, they showed that even nonsense reasons do remarkably well at increasing acceptance rates of small favours. The authors attribute this phenomenon to ‘mindlessness’. In this context, we can consider mindless acceptance of a request as a simple type 1 response, associating a reason with necessary compliance. Importantly, they also demonstrated that for larger favours, the required reasoning offered had to be sufficiently more robust in proportion to the ask, which itself suggests a system 2 response has been elicited, as the results do suggest.
So how did this study work? In the classic experiment, a researcher would hang around the Harvard University library photocopier until they saw it in use and a queue developing. At that point, the researcher would attempt to cut to the front of the queue. This would not go down well in the UK one imagines, but the experimenter was prepared with three variations of a question to ask. The small ask version went something as follows:
- Version 1 (request only): “Excuse me, I have 5 pages. May I use the Xerox machine?”
- Version 2 (request with a real reason): “Excuse me, I have 5 pages. May I use the xerox machine, because I'm in a rush?”
- Version 3 (request with a fake reason): “Excuse me, I have 5 pages. May I use the xerox machine, because I have to make copies?”
The success of the experimenter may surprise you. When the data was analysed, the following was found:
- Version 1 (request only): 60 percent of people let the researcher skip the line.
- Version 2 (request with a real reason): 94 percent of people let the researcher skip ahead in line.
- Version 3 (request with a fake reason): 93 percent of people let the researcher skip ahead in line.
Results
The data shows that reasons drive the success of small favours and requests. Even a nonsense reason prompts a significant increase (33%!) in acceptance. And whilst it may be tempting to put the success of the researcher down to a social pressure to avoid confrontation, which may explain the reasonably high success rate of V1 - we can’t infer that from the data. What we should take away from this research is the importance of the word ‘because’ when you’re trying to influence customers and other people around us to do small favours.
In the case of small favours, as recently as 2009 this study has been effectively replicated producing almost identical results – see Key, Edlund, Sagarin and Bizer[3]. The suggestion is clear: people are biased towards accepting small favours on face value if given a reason, and this heuristic can be used to great effect in eliciting compliance in small asks.
But there’s more! Remember, Langer et al found that as the demands of the favour increased, so did the informational requirements of the reason:
The table from the original study shows that when the ask was larger – when the researcher asked to copy 20 pages rather than 5 – the acceptance rate was dependent on a sufficiently detailed reason.
Providing sufficient information almost doubles the success rate of giving no reason or a nonsense one (18% increase in acceptance!). In the larger favour example, there is a clear difference in the tolerance of placebic vs sufficient info. This is very likely due to the activation of system 2 given the greater ask, and therefore the greater likelihood of thorough evaluation of supporting reasons.
Reasons in the customer journey
In most of our client’s customer journeys, we are trying to move customers towards one thing or another. It might be encouraging a customer to troubleshoot problems with their home broadband services or encouraging a customer to return their telecommunications equipment. We know that reasons matter to our client’s customers, so when we make attempts to influence them, we make sure to provide valid reasons to promote response and success rates. We’re also aware of the ask we’re making – so we take care to make sure that the reason we give has sufficient informational content and justification, proportionate to the favour we are trying to elicit.
You may be able to implement reason-based directives into your customer journey right now. Are you calling customers to action without a reason? Or maybe you’re taking for granted that the customer is motivated enough? We have found that even if logically one would expect a customer to be sufficiently motivated, a carefully designed and directive prompt may be necessary to improve success rates within a given journey.
We manage the hurdles of short format, text-based conversations by using direct, reasoned instructions. This helps initiate a system 1 response, and you can do the same by keeping your language directive and offering sufficient reasons for actions. Is this something you’re doing in your current customer journey? It’s important to ask these types of questions of your own processes to avoid missing out, because providing appropriate nudges often positively affects CX and success rates.
What’s next
It just takes some concentrated effort, trying to understand the context of your customers' decisions and making the appropriate linguistic adjustments. The benefits are tangible – supported by a conversational AI, our proactive customer journeys save our customers millions of pounds and dollars every year. If you’re interested, you can find out more here.
Now that you’ve got an idea of the science and theory behind nudges, you’re in a position to start designing and implementing them! In the next post, we put it all together. We discuss wise interventions – designing nudges for the public good and the practical considerations to keep in mind. You’ll learn the principle behind an effective design, what reporting considerations to keep in mind and practical details to make your nudges successful.
