Misunderstandings

Front matter

Be his >

MISUNDERSTANDINGS FALSE BELIEFS IN COMMUNICATION

F. Saree” te a E

Contents

1 9 17 31 5 Listening 47

= N w

Perceiving how they tal Perceiving how they listen 89 In higher order: Seeing their view of our view 101

Chapter 1 Introduction: Should we talk”

Three tales that feature in this book

The governor exudes power. Rachel watches him from across the conference table. His occasional laughter is loud, his knowing smile invites others to speak, but not for too long. Today is his first visit to the institute. The institute’s director has introduced the senior researchers sitting around the table, Rachel among them, and two of her colleagues have already made their contributions to the conversation. “Who else?” says the governor.

Dimitri finds Agniezka at the elevator. They are on their way back to their offices, returning from an inter-group competition where they jointly presented a product — and suffered a defeat against a group of colleagues who showed a more effective design. Dimitri and Agniezka had worked on their idea for about a year, with high hopes to win today and move to the next stages of development. Now, they have to reconsider. The elevator door opens.

Steve carries his ball in a swift walk around the block. He is looking for other kids to play with. Turning a corner, he almost bumps into Ralph. Ralph is much older, at least nine, and he is one of the schoolyard bullies. Today, on the sidewalk, he does not look bully-ish at all. His shoulders are drooping, his facial expression is sad and he leans against the wall of a building, as if seeking its support. The street is empty and Steve can see that Ralph was alone before he, Steve, had turned the corner.

© 2023 Georg Weizsäcker, CC BY-SA 4.0 https://doi.org/10.11647/OBP.0367.01

About the book

Should we talk? This book treats this question, and many others, as being empirical in nature: if we were to live through a given situation in 1000 parallel lives, where in one half we talked and in the other half we didn’t talk, in which of the halves would we fare better?

The results of such a thought experiment vary between Rachel, Dimitri, and Steve — or any other conversation that we may experience. Costs and benefits depend on the details of the conversation. Every conversation is different and rich in detail. Why, then, is this book so thin? The main reason is that the book, while asking questions with the usual scholarly care, is quite brief in its way of answering these questions. More precisely, it contains one empirical answer per question — appearing in separate font - and this answer is valid only for a specific context. The reader is invited to extrapolate, asking the same question for other conversations of interest.

Such extrapolation is, luckily, easy. Having conversations is an ubiquitous experience. Countless literatures give examples of noteworthy conversations, in the social sciences, business studies, psychology, and many other fields.

While the answers in this book are few, the set of questions is suprisingly exhaustive. The book’s 18 questions about human conversations are, in a specific sense, all relevant questions up to a well-defined point. They systematically cover the ways in which a conversation can go wrong.

So, what are these ways of going wrong”? This book describes a particular set: inaccurate expectations. We may fail to judge the situation correctly. We may think that things are plausible when they are not, in a way that reduces our utility. We may misjudge our partner in conversation, or the situation, or the language. The book thus approaches conversations through the expectations that they come with. It formulates 18 misunderstandings: ways in which expectations are off target. Misunderstandings may be the ugly ducklings within the family of beliefs. But hey, here they are. Sometimes, they may even be more important than their not-so-ugly cousins. It is therefore worthwhile to study their nature.

This leads to asking about the reader’s benefit — why read this book? The main benefit is, perhaps, to learn something new in behavioral economics. Communication is a fast-growing research area in economics, and especially in microeconomics and organizational economics. Experimental measurements

1. Introduction: Should We Talk? 3 of communication abound, too. But misunderstandings, despite appearing very frequently, are an uncharted area of research. Little is known about them, at least in the scholarly literatures of economics and business studies.

To help fill this gap, the book aims to give an overview of misunderstandings. It discusses the events, statements and actions that the misunderstandings are about — the “targets” of beliefs that are relevant in a conversation. In parallel, the book discusses the methods to measure the beliefs and it reports evidence on the distance between the beliefs and their targets. All of this follows an empirical approach. Each of the 18 questions put forward here is an empirical question, i.e., in principle its answer can be found through a well-designed measurement.

A notable distinction between this book and the descriptions of most other non-standard beliefs in behavioral economics is that misunderstandings are not described here with reference to a theoretical benchmark. Instead, a misunderstanding is just a belief that misses its target, empirically speaking. The book uses its conceptual parts merely as a springboard.

The non-theoretical exposition also means that the reader does not have to go through elaborate economic models. Having a general understanding of statistics suffices to follow the analysis. As a note of caution, however, the reader should know that parts of the analysis, especially in the book’s second half, are game-theoretic in nature, and thus non-trivial.

Credit where credit is due: psychology scholars have a long and valuable tradition of documenting misunderstandings and, more generally, all kinds of non-optimal beliefs and behaviors. The book includes some of these findings in its discussion. Interdisciplinarity is, always, key in behavioral economics.

The second benefit for the reader also has an interdisciplinary flavor: he or she will learn about the logic of human language. The book ventures outside of both psychology and economics: to discuss false beliefs in communication, one needs to ask basic questions about the interpretation of messages. Such questions are the object of study in the field of pragmatics, a highly developed part of linguistics and the philosophy of language. This field, too, has a tradition of studying misunderstandings, and significant parts of the book’s analysis rely on insights from it.

The author hastens to add that the book is not a suitable introduction to any part of psychology or pragmatics. It merely makes observations about these fields and adapts some of their basic ideas from the vantage point of an economist. The main observation about pragmatics is a simple one: many of its ideas can be suitably studied via a description of beliefs. Economists are used to studying beliefs — but the beliefs that pragmatists emphasize are largely new to economics.

A well-known economic anecdote may serve as illustration. In July 2012, in the midst of the euro crisis, the European Central Bank’s president Mario Draghi made history with a short utterance: “Within our mandate, the ECB will do whatever it takes to save the euro. And believe me, it will be enough.” The statement was an instant hit — media reported it widely and the statement is now believed to have caused a market turn-around: speculation against the government finances of multiple countries, and against the euro, subsided. Markets returned to normal.

How can a simple statement be so influential? Or, expressed in terms of beliefs: What did the listeners expect when they heard it? What did Draghi expect when he said it?

A large part of the answer is linguistic. The statement carries much meaning and strength by alluding to several important contexts at once. First, Draghi points out that the scope of ECB’s activities is wide (“whatever it takes”). This was understood to indicate that the ECB would buy government bonds at a large scale if necessary. Second, he points to the depth of the bank’s pockets (“it will be enough”). Third, he points to the fact that he is a credible source of information about the previous two contexts and that the market traders, and all other listeners, know less about them (“believe me”).

The third of these contexts is especially relevant: presidents of central banks have an almost hard-wired commitment to being credible. Draghi ties this credibility to his knowledge about what the ECB can do. This is the context that he wants the listener to think about. A context where he, Draghi, is the master in the ring, which gives the statement its strength: the listener has no alternative but to believe him.

A feature of his statement, one that it shares with other successful statements, is that the listeners understand these relevant contexts although Draghi only ever-so-slightly alludes to them. But how? In a complicated world with a vast set of candidate contexts, how do the listeners learn what the talker refers to? It is a little miracle when it works.

1. Introduction: Should We Talk? 5

A second example, also from the policy world, describes a misunderstanding. In April 2021, Germany’s political parties were in an open race for the succession of Angela Merkel as chancellor. Much of the political debate focused on Covid, at a time when the roll-out of vaccinations was still slow and a new wave of infections swept across Germany. Armin Laschet, a party leader and a front-running candidate for the chancellorship, made a public statement on Covid regulations. It was widely followed by the media, not least because Laschet himself had built up suspense by pre-announcing the statement a few days earlier. Then he finally made it: “We need a bridge lockdown.” The idea being, as he explained, that another temporary restriction of public life, while costly, may serve as a bridge to better times: the lockdown would give time to create more vaccine-induced immunity in the population. However, Laschet failed to gain support for his initiative. Instead, large parts of the public criticized him for inventing the new expression “bridge lockdown”. It was argued that using the positive-sounding word “bridge” to counter-act the public’s negative emotions about the restrictive measures amounted to being borderline deceitful. Also, commentators noted that all Covid lockdowns have the nature of serving as bridges to better times, such that Laschet said little that was new. (He later also lost the election.)

Laschet suffered from a basic misperception: he believed that the new, better-sounding wording (with an unusual sound also in the original German, “Brücken-Lockdown”) would make the public more favorable to his position. Alas, it did not. The choice of words drew attention away from the statement’s substance — which was not generally viewed as unreasonable — to the more superficial topic of the intention of the politician who uttered it. The statement lost its meaning by revealing that Laschet wanted it to sound meaningful. With such a strong emphasis, the public wanted more content than Laschet offered.

The later chapters of the book will discuss how such linguistic mechanisms may, or may not, work in general. As a one-sentence preview of this discussion, they may work because the listeners can see the conversation through the eyes of the talker. They interpret what the talker intends to achieve with the statement and what context the talker refers to. They search for the relevance of intentions and contexts, guided by the statement to find a useful insight (in Draghi’s case) or not (in Laschet’s case).

The reader may already recognize the sound of game theory here: a listener who thinks strategically does not only solve his or her own problem, but solves the talker’s problem, too. Such second-order thinking is quite a task in a conversation because the listener needs to do several things at once: anticipate the preferences of the talker and consider the context that the talker refers to, among the many possible contexts that exist. It is complicated, and an invitation for misunderstandings, but not impossible.

Related questions may spring to the reader’s mind, too. Doesn’t the talker do the same, and solve the listener’s problem? If so, don’t both people need to engage in another layer of iterated strategic thinking, where each thinks how the other person, conversely, thinks about them? And so on — how many layers are there?

The book discusses conversations, by and large, with only two layers of strategic thinking. It sets out the definitions of the first-order beliefs that are relevant for the talker and the listener, then discusses these beliefs at length, then turns to the other person’s expectations about them (i.e., second-order beliefs). As Chapters 9 and 10 will argue, the analysis could go further and consider beliefs of third and higher order, but they also argue that such a higher-order analysis would perhaps be less valuable than the analysis of lower-order thinking. Beliefs of higher order become less and less easily measurable, and studying them is often not necessary: many key insights arise without reference to higher-order thinking.

Chapter 11 gives some leads to further readings. The chapter also explains where each of the substantial ideas used in this book originate from. (This is important because the book’s main body contains no bibliographic references. The chapter serves to give appropriate credit to other authors and literature, in the same way that other books or articles have bibliographic endnotes. )

Readers of all backgrounds will find that the book is short and dense. It is best to read it slowly. Or, perhaps, to complement its reading with that of other texts. The book is, however, a stand-alone, self-sufficient piece.

A note to course instructors who consider teaching with the aid of this book: it’s best if you judge for yourself whether it works for your course. The author has repeatedly test-taught the book as the main reading for a Master’s-level student seminar on the behavioral economics of communication. Students of different backgrounds can deal with the interdiciplinary material in very different ways, going deeper into secondary readings accord-

1. Introduction: Should We Talk? 7 ing to their own interests. The expositional style of asking questions aims to teach students about empirical research in a simple and transparent way. Finally, note that many different ways of splitting up the material are possible, making the book suitable reading for a half-term course or a full-term course.

The book’s structure follows a rigorous but unconventional logic. The next chapter lays the groundwork and describes the book’s assumptions about conversations. Essentially, it describes an analytical framework based on subjective expected utility theory (which implies large degrees of freedom). The subsequent six chapters formulate and discuss the 18 misunderstandings, three per chapter. Each is presented as a question about a possible misunderstanding, digging deeper as the book progresses. Chapters 3, 4, and 5 cover possible failures of first-order beliefs, and Chapters 6, 7, and 8 cover analogous failures of second-order beliefs. Chapters 9, 10, and 11 discuss how one may find more possible misunderstandings and how else the 18 questions can be placed in the wider literature on human communication. These three chapters are not part of the book’s main body (reading them is very much optional) but they hopefully add clarity. The book’s focus on misunderstandings is, definitely, not an excuse for avoidable confusion.

As the reader may have guessed already: the examples involving Rachel, Dimitri and Steve, which will re-appear throughout the book, are fictional and only serve for illustration. The evidence from scholary research, in contrast, is real. It is only a thin layer of evidence but appears to show a pattern: many beliefs in communication are not very sophisticated in that they do not react to one or several pieces of information to a sufficient degree. Either the talker or the listener fails to take something into account.

Future measurements will tell us about this pattern’s robustness. As described above, the book is written at a relatively early point in the sequence of the relevant research that it describes. It is much more about questions than about answers.

Chapter 2 Having a conversation

Three conversations

Agniezka and Dimitri stand in silence as the elevator moves up one floor, and another one. Dimitri then addresses the issue of the lost competition: “This was bloody. Not the end of the story, though. We will hit back.”

Agniezka turns her head and gives Dimitri an expressionless stare. She says, “Hopeless. Make it your bloodbath, not mine. I’m done with this project, sorry.” The door opens and Agniezka leaves the elevator.

Rachel straightens her back. “Mr. Governor, please consider the urgency of research on stolen artifacts. We ask you to look into the possibility of funding a new center of competence that studies the provenance of cultural property. How did this property leave its home? The Cambodian statues in western museums and collections are a good example. Many of them were looted in postwar turmoil. This was publicized in newspapers and we verified the reports with scientific studies. But for contexts outside of Cambodia, the facts about provenance are mostly unresearched. We need more structured knowledge, we have enormous work before us and we need your help with it.”

The governor hears her out and replies, “Rachel, another great line of research! We are on the same page and I much admire your spirit and that of your colleagues. I well recall how you and I first met, in a panel discussion on cultural policies, planning a common agenda. Let’s continue this process! You also know that we already do a lot for top-level research in our museums, through our funding programs.” He pauses briefly. “We will continue these efforts, too — but here and now, we have to move on.” He looks at the

© 2023 Georg Weizsacker, CC BY-SA 4.0 https: //doi.org/10.11647/OBP.0367.02 researcher sitting next to Rachel.

For a few seconds Steve looks at Ralph, the older boy, who turns around and looks back at Steve. His shoulders are not drooping any more.

“Wanna play?”, says Steve, holding the ball in his hands.

Ralph’s face assumes its usual, hostile look. He walks straight towards Steve and slowly, almost routinely, pushes him with his overweight arms, then walks away.

Simplifying restrictions

A conversation has multi-faceted content, much more than meets the eye. The three conversations of Rachel, Dimitri and Steve are similar only at the surface — the talker wants the listener to do something, the listener rejects it — but they strongly differ in terms of their contexts, their human characters, their languages, their outcomes, and in the future events that they induce.

The analysis starts by narrowing things down. It focuses on what is uncertain. When people talk and listen, they do this with incomplete knowledge. Three aspects of uncertainty arise: about actions, issues, and types.

Actions are everything that people do during or after the conversation. This includes all their possible statements: statements are actions.

To notate the action (or statement) of person i, the book uses the symbol a’ and it is understood that the action is one out of a given set A? of possible actions for person i, i.e., a’ € A’.

The book will keep this and other mathematical notations to a minimum and it sometimes even describes variants of the basic set-up without notating them formally. For example, a person’s action may sometimes depend on another person’s action without a mathematical notation for it.

Issues are things that are exogenous to the actions. They are things that cannot be changed. The book refers to them as the “state of the world” (or simply “state”), which is notated by w and is, generally speaking, unknown

2. Having a Conversation 11 at the time of the conversation. The true state is an element of a larger set of possible states, w € Q.

Not all is unknown of course, as everyone knows something about the state of the world. The information that person 7 has about w is notated by the symbol J’. It is convenient to describe I’, as a subset of Q that contains w; this indicates that person ʻi, if she pays attention to her information, can deduce that some states are ruled out: those that lie in Q but not in J’. She does not know the exact value of w but she can safely conclude (later, we will say, it is “manifest” to her) that w lies in I.

Most likely, the people in the conversation each know something different. The conversation may help to exchange their views.

Types, notated by 6° for person i, describe the person’s preferences. They are person-specific and include everything that is idiosyncratic: desires and tastes, capabilities and necessities.

Note that a person’s preference for certain outcomes can be related to her behavioral constraints. If person 7 strongly dislikes an outcome, this is similar to the case where her action set A’ does not contain an action that leads to this outcome. The book’s analysis keeps these things apart, for simplicity: 6’ describes person i’s preferences for a fixed set A’, i.e., it describes which of the available possible outcomes (and actions) she likes relatively better than others.

In its general discussion, appearing in normal font and black color, the book does not specify the exact nature of actions, issues and types. In any given conversation, they may be either explicit and in the open, or hidden in the background. For a successful analysis of a given situation, one should specify these aspects in sufficient detail but also simplify where appropriate, focusing on the main features of the situation.

At the surface, Steve’s simple question (“Wanna play?”) asks about Ralph’s preference regarding a ball game. Ralph’s possible reactions, however, are not all playful: he may turn to violent bullying, especially now that Steve saw him in a state of despair. The analysis may therefore collect Steve’s possible statements into two groups: those that address Ralph’s state of despair, e.g. by asking about it, versus those that address more normal things. We also separate the set of Ralph’s possible reactions in two groups, describing whether he is violent or not, and likewise, his set of possible types:

his despair may either fuel his tendency to be violent, or it may reduce it. Steve’s own type, in contrast, plays no important role in this conversation. The relevant issues, or states of the world, are the unknown events that underlie Ralph’s despair. A further relevant unknown are Steve’s future actions: will he tell other kids about what he saw, or not?

Rachel focuses her statement on a set of future actions that she hopes to provoke — the governor’s possible support for a new research center. She could talk about many different things but here, too, we may simplify by restricting her choice to be between two topics: the process on cultural policy taht has already begun (and that the governor refers to in his response) or the new topic of provenance. We may also describe the governor’s possible reactions as binary: either supporting Rachel’s research with new funds, or not supporting it. We note that the relevant state of the world comprises not only the substantive topic of conversation but also the professional context, i.e., the unusual situation of the governor’s visit as well as the presence of other colleagues. We also note that the personalities, or types, of Rachel and the governor are in the background; no-one addresses them openly — only the personal tone of the governor’s response alludes to them.

Dimitri is hoping to keep Agniezka on his team and to jointly prepare an aggressive push in the development of their project. The circumstances of the conversation are fairly complex but for our discussion we again focus on a small set of issues: the team’s chance of success if the team stays together, the team’s chance of success if the team splits up, and Dimitri’s and Agniezka’s history of past actions. This history is in the background but influences the conversation, and more elaboration on it will follow in later chapters. We can also restrict attention to two types of statements that Dimitri may make in the elevator: statements about the possible scenario where Agniezka remains on the team — he indeed chooses to make such a statement — versus statements about the possible scenario where the team splits up. In either case, his statement is bound to be fairly bad-tempered, owing to his disappointment about the lost competition. Agniezka’s reaction, in turn, can be simplified to either staying on the team, or not. (Notice also that Dimitri does not have to say anything during the elevator ride. We observe that this may be the wiser course of action, but the subsequent discussions will take it as given that Dimitri is the talker.) The relevant types are Dimitri’s and Agniezka’s preferences vis-a-vis these possible statements and vis-a-vis the future actions that may arise in each of the possible scenarios.

The book makes additional restrictions, to narrow things down further.

2. Having a Conversation 13

The first restriction is to consider only single steps of a conversation: there are two people in the conversation — the two interlocutors — where one person says something and the other person reacts.

One may try to apply the book’s analysis to longer conversations with more than one step. To do so, one could simply view everything that is said in the conversation as a single statement. The analogy is vague, however, and it is therefore best to focus on single-step conversations.

Future actions of the interlocutors may, of course, have a foreshadowing effect on the present conversation. For example, the possibility of telling other children about the encounter with Ralph is on Steve’s mind. The book does not analyze how the interlocutors choose such future action — they are not part of the present situation. It merely considers that future actions are plausibly influenced by what is said now. For a streamlined notation, all future actions are subsumed as parts of the interlocutors’ action sets, even though the interlocutors cannot pick them in the present conversation.

As a second restriction, the book assumes that each person knows their own type. Introspection is assumed as faultless. In contrast, the other person is somewhat mysterious. In formal notation, let I}, be person i’s information about person j’s type, and let ©) be the set of possible types for 7. Then, the above assumption is that I}, = 0' and Ij, C ©”.

The third restriction is that the uncertainty about other people’s types is assumed to be statistically independent of the uncertainty about the state of the world. This means that one cannot learn anything about the state from learning about the type, and vice versa. It is another simplification; in its absence the exposition would be cumbersome, with little insight added.

Translating this assumption into a formality is not possible yet, as we have not introduced any formalities about probabilities and beliefs. Once they are introduced, further down in this chapter and the subsequent ones, making the assumption of statistical independence is straightforward. The book’s analysis also assumes that the interlocutors agree to this simplification.

More simplification comes from the fourth restriction: person 2’s realized utility from the conversation depends only on the realized values of the uncertain aspects. These are: the actions of the two interlocutors (a’, a’) (including their statements in the conversation and everything they do afterwards), the realized value of the state of the world w, and the person’s type 0".

The fifth restriction is perhaps the most controversial: everyone deals with the uncertainty in a subjectively “rational”, expectation-based way. That is, while the people in the conversation cannot know the uncertain aspects when having the conversation, they form subjective expectations about them and react to these expectations.

Formally, the fourth and fifth restrictions amount to assuming that person is utility from the conversation is given by ut: Ax AIxQXx0O' oR

(which means that u is a function with arguments a’, af, w, and 0°) and that the chosen action a’ maximizes i’s subjective expected utility, i.e., she acts to maximize the expected value of u’ given her subjective expectations.

The book’s focus is on these subjective expectations. For brevity, call them beliefs. The book simply asks, “What are the beliefs that would justify leading the conversation in the way that people lead it?”

The reader may wonder, why ask such an indirect question about underlying beliefs? Why not ask questions about the conversation itself?

The answer is twofold. First, the book can rely on the existing literatures on decision theory and game theory. These are impressive bodies of literature that describe, among other things, the existence and other properties of beliefs that justify actions. Without going into detail, it is safe to say that the book solidly stands on the shoulders of giants.

Second, a key observation lies at the heart of the book: conversations cannot be correct or false in an unambiguous way, but beliefs can be correct or false. Moreover, one can measure whether they are correct or false.

The next chapters therefore ask whether the beliefs about the uncertain aspects are distorted. For formal notation, the book uses the symbol PX for person 2’s subjective belief about unknown item x. It is a probability distribution over x’s possible values and will depend on information that person i has. For example, i’s belief about the state of the world after seeing her information set I’, is P/(-|I3), a probability distribution over the elements in Q. As another example, P/,(-|I},) is i’s belief about j’s action af, after

2. Having a Conversation 15 learning information J}; about j’s type, and it is a probability distribution over the elements of A’.

This leads to the final simplifying restriction: common knowledge of the conversation’s basic ingredients. The sets of possible actions, types, and states of the world, are assumed to be commonly known by the interlocutors and the information structure that governs who may receive what information is also assumed to be commonly known by them. That is, while each interlocutor does not know what the other knows, they do agree on what this knowledge could conceivably be. Moreover, the fact that each interlocutor reacts to their belief, as specified by subjective expected utility, is also taken as commonly known. The interlocutors do not know each other’s beliefs, but know the fact that the other interlocutor reacts to some beliefs.