THE CORE QUESTIONS AND RESULTS OF THE DEMOCRATIC DILEMMA

The subtitle of our 1998 book asked, “Can Citizens Learn What They Need to Know” so that, in turn, they can perform their duties within their democracy. Citizens are asked to be overseers of democracy. They are asked to serve as voters, jurors, board and council members, and legislators. A principal attack on democracy, especially in America, is that citizens cannot attain the knowledge they need to perform these duties and make democracy work.

In The Democratic Dilemma (Lupia and McCubbins Reference Lupia and McCubbins1998), we sought to clarify conditions under which citizens who had limited information could make the same choices they would have made with more information. The logical foundation of our work was a series of extensions of Crawford and Sobel’s (Reference Crawford and Sobel1982) classic “cheap-talk” communication model. We used these extensions to derive conditions for trust, persuasion, and effective learning in a range of decision-making contexts.

Our work revealed four sufficient conditions for trust in these contexts. Specifically, when attempting to derive an accurate inference from what someone else says, a person must believe that (1) she shares with the speaker common interests in the outcome of their communicative interaction; or (2) the speaker paid a sufficiently large cost just to speak with her; or (3) the speaker’s statements are subject to sufficient threat of external (third-party) verification; or (4) a speaker is subject to sufficiently large (third-party) penalties for lying. Various combinations of these factors also are sufficient. The listener will trust what the speaker says when any of these conditions are upheld.

If the listener also correctly believes the speaker’s statement to contain relevant expertise that she does not already hold, then the consequence of the communicative act will be a persuasive statement that may lead the listener to have more accurate beliefs about the decision in question (depending on whether the speaker has the needed expertise). If this increase in accuracy is sufficiently large, then we can say that the communicative act increased the listener’s competence in making the decision.

From this logic, we argued that reasoned choice (i.e., the choice that a listener would make if she had the best available information) does not require actually holding encyclopedic knowledge about a decision. Rather, it merely requires that a person be able to identify reliable sources of relatively simple pieces of information that lead her to the same decisions that she would have made if she knew more.

An implication of these findings is that institutions—such as rules of evidence in courts and agencies as well as legislative rules and the competition between parties and candidates for office—can help people make better decisions about which information to trust and which to ignore. This outcome occurs when institutions generate one or more of the conditions described previously (something that Facebook and other social media platforms still struggle to create).

The results on trust and persuasion also allowed us to model problems of delegation. Delegation is ubiquitous. We can think abstractly about all delegations as an “agency problem.” The principal is the person who requires a task to be performed; the agent is the person to whom the principal delegates authority to complete that task. Redelegation also is ubiquitous. However, there are always perils from delegating, in the form of agency losses and agency costs. The former, agency losses, are the principal’s welfare losses due to the agent’s choices when those choices are suboptimal from the principal’s perspective. The latter, agency costs, are the costs of managing and overseeing an agent’s actions (e.g., the agent’s salary).

Three necessary conditions give rise to agency losses and, thus, the delegation dilemma. The first condition is that the agent must have agenda control. That is, the principal delegates to the agent the authority to take action without requiring the principal’s consent in advance. This puts the principal in the position of having to respond to the action ex post rather than being able to veto it ex ante. The second condition that underpins the delegation dilemma is that there will almost always exist a conflict of interest between the principal and the agent. If the two have the same interests, or if they share at least some common goals, then the agent likely will choose an outcome that the principal finds satisfactory. If not, then the third condition is that the principal must be unable to effectively check an agent’s actions. Conventionally, the lack of an effective check is attributed to the agent’s expertise—that is, the agent may be chosen because of this expertise in the first place, but the indirect consequence is that the principal may be unable to evaluate the outcome of the agent’s choice. In The Democratic Dilemma, we presented the necessary and sufficient conditions for solving the delegation dilemma, showing when principals can delegate and be made better off by their agent’s actions.

Bureaucratic expertise relative to members of Congress, for example, is an often-discussed reason that delegation to the bureaucracy becomes unaccountable and unsuccessful. However, the problem is not that legislators lack information or that bureaucrats monopolize information; rather, it is how principals are to assess the accuracy of the information they receive. This might seem to imply that to ascertain whether an agency is doing its job, political leaders must engage in proactive oversight: members of Congress, senators, and the president must gather sufficient information, before the agency’s actions, to assess whether it is pursuing its mission in a way that improves the welfare of political leaders.^{Footnote 1} To do this, political leaders collect and correlate enough information to make accurate inferences and then reach reasonable conclusions about whether an agency is serving their interests. In chapter 5 of The Democratic Dilemma, we identified the conditions under which conflicting interests and information asymmetry between the principal and agent cause delegation to succeed or fail. We proved that if a principal has access to the testimony of others and either the ability to identify testimony from a knowledgeable and trustworthy source or access to institutions that give the principal this ability, then delegation can succeed even if the dilemmas of delegation are present.

A final contribution of The Democratic Dilemma pertains to attention. Our work demonstrated that a person’s expectations of learning the truth, as opposed to learning a falsehood, influence attention-related incentives. In turn, conditions for trust influence how a person distributes her limited attentive capacity.

SUBSEQUENT INNOVATIONS

As a whole, The Democratic Dilemma clarified how people and institutions can—and cannot—reconcile democratic responsibilities with cognitive limitations. This work also generated a series of notable inquiries into how the presence of multiple information sources affect trust and learning outcomes. For example, Boudreau and McCubbins (Reference Boudreau and McCubbins2008) built on our experimental design to show that competition between experts is not sufficient to generate conditions for trust, persuasion, and learning. Rather, competition can have this effect only when conditions for trust and persuasion are present in sufficient quantities to generate truthful statements and associated knowledge acquisition. Boudreau and McCubbins (Reference Boudreau and McCubbins2010) tested whether a “wisdom-of-crowds” effect can substitute for the conditions for trust and persuasion. In these experiments, Boudreau and McCubbins offered experimental subjects a choice between a poll of the choices that all subjects recommended versus a clear statement made by speakers who were known to listeners to satisfy the conditions for trust and persuasion. Boudreau and McCubbins’s results showed the power of the wisdom of the crowd as well as its folly. In their work, unsophisticated subjects were more likely to rely on polls than trustworthy experts when the choices were difficult. The folly arises from the fact that, when the problems are most difficult, a poll across a group that includes many nonexperts will almost always be uninformative or wrong. Sophisticated subjects are more likely to solve the problem on their own. Hence, sophisticated subjects rely less on polls and are rarely misled. Together, these experiments—along with complementary efforts by Boudreau and McCubbins (Reference Boudreau and McCubbins2009) and Boudreau, McCubbins, and Coulson (Reference Boudreau, McCubbins and Coulson2008)—reinforced and extended our (1998) insights about how contextual changes can facilitate reasoned choice.

Other work focused on how attributes of communication networks facilitate trust and reasoned choice. McCubbins, Paturi, and Weller (Reference McCubbins, Paturi and Weller2009) and Enemark, McCubbins, and Weller (Reference Enemark, McCubbins and Weller2014) experimentally examined the role of different types of connections within communicative networks. These authors argued that the primary function of connections in a network is to disseminate information. They showed how giving an individual more information about a network’s structure can have the same types of effects as adding additional network connections. They demonstrated that if the information produces the conditions for trust described previously, it can help subjects coordinate and accomplish tasks more effectively.

Thinking further about the types of information that can facilitate reasoned choice, Burnett and McCubbins (Reference Burnett and McCubbins2013) and Burnett, Garrett, and McCubbins (Reference Burnett, Garrett and McCubbins2010) extended Lupia’s (Reference Lupia1994) emphasis that voters can use endorsements to make reasoned choices in initiatives and referenda. In each study, they conducted surveys during these types of elections. They showed that the endorsements have salubrious effects only when Lupia and McCubbins’s conditions are satisfied—that is, when a voter believes an endorser is knowledgeable and that the endorser is perceived by the voter as trustworthy.

Following publication of The Democratic Dilemma, the results that we described were replicated and extended into more specified models of learning, voting, and decision making. Taken together, this literature has made important contributions to how scholars and practitioners understand what information is and is not sufficient for reasoned choice.

A SURPRISING SUBSTANTIVE IMPLICATION

The world has changed in important ways since publication of The Democratic Dilemma. One change is the amount of information available on a wide range of topics. When the book was written, it was fashionable to conjecture that a lack of information was the main problem facing voters. Today, the main concern has a different character. Misinformation and fake news combined with a better understanding of motivated reasoning has prompted most present-day observers to be concerned not about a lack of information but rather about how citizens will sort through and use all of the information that is available.

Across the world, the lowered cost of accessing information and the proliferation of different kinds of information has produced crises of credibility. Formerly trusted sources of information ranging from the mass media to government are now questioned more frequently—and, in many cases, more fervently—than ever before. Science has not been immune from these changes. On high-profile issues such as climate change, genetically modified organisms, and vaccination, substantial questions about the validity and value of scientific research have been raised.

In response to these challenges, many science organizations have pursued ways to communicate more effectively. This is a significant change. When The Democratic Dilemma was written, and for decades prior to its publication, professional incentives for most scientists did not include the ability to communicate effectively with broad audiences. Instead, the main currencies of the scholarly ecosystem that fueled academic careers were publishing in academic journals, presenting papers at academic conferences, and obtaining funding from science-focused organizations. To achieve these goals, the communicative skill needed was to convey insight and value to other scholars who often shared detailed substantive and methodological affinities. For many scholars, the “ideal reviewer” was one who appreciated the same jargon and conceptual frameworks as they did.

Today, most people obtain information about science through electronic devices—the same devices that show funny movies, live sporting events, memes, video games, social network connections, and so much more. Hence, for scientists to communicate effectively, they must compete in a fierce battle for attention (Lupia Reference Lupia2013; Reference Lupia2016). The Democratic Dilemma’s model of attention laid out fundamental parameters of this battle. The model’s initial premises drew from literatures in psychology and neuroscience that, at the time, documented extraordinary limits on human attentive capacity. These limits produce strong incentives to direct attention to high-consequential stimuli. Subsequent research in brain-related fields showed how human cognitive systems actively seek to hide attention limits from consciousness during real-time information processing (see, e.g., the review in Peterson and Posner Reference Peterson and Posner2012). An implication of these findings is that we would not expect untrained science communicators to be aware of these limits in themselves or in others. The Democratic Dilemma provided an early theoretical template for thinking about communication strategies in such circumstances.

Processing scientific information, however, requires more than obtaining attention. It requires keeping attention for a long enough time to form new memories. To understand these phenomena, and to help a broader range of researchers and science organizations produce content from which people will want to learn, a science of science communication has emerged (see, e.g., Jamieson, Kahan, and Scheufele Reference Jamieson, Kahan and Scheufele2017; Suhay and Druckman Reference Suhay and Druckman2015). Key insights in this literature relate to how to obtain trust and credibility (for reviews of foundational literatures see, e.g., Chong and Druckman Reference Chong and Druckman2007; Druckman and Lupia Reference Druckman and Lupia2000; Reference Druckman and Lupia2016). The Democratic Dilemma’s emphasis on conditions for trust provides the explicit theoretical foundations for this work. At a time when many scientists struggle with the premise that their credibility is not preordained but rather the result of knowable theoretical conditions, The Democratic Dilemma clarifies ways in which science communicators can effectively offer important information.

Collectively, this scholarship has had a substantial effect on practice. Organizations such as Climate Central—whose content is used by news organizations around the world—and the National Academies of Science, Engineering, and Medicine’s “From Research to Reward” educational series are drawn explicitly from the lessons of this research. Whereas many hurdles to effective science communication exist, literature that built from The Democratic Dilemma’s foundation are enabling more scholars to help more citizens make reasoned choices about more topics than ever before.