Category Archives: epistemic networks

Mapping Ideologies as Networks of Ideas

Newly published (behind a firewall): Peter Levine (2022) Mapping ideologies as networks of ideas, Journal of Political Ideologies, DOI: 10.1080/13569317.2022.2138293.

This is an early publication from my main current scholarly project. I am concerned that several streams of research and intellectual conversations are converging on the same conclusion: people just aren’t very thoughtful or rational about politics. This stance discourages efforts to enhance democracy and the public sphere. However, the most prevalent measurement tools–standard opinion surveys–are systematically biased against detecting the complexity and individuality of individuals’ political views. Various colleagues and I are experimenting with alternative quantitative methods that involve directly asking people about the connections among their distinct beliefs and analyzing the results as networks. This sole-authored article is a pilot study aimed at validating the method.


Individuals in a non-representative sample of 93 US progressives were asked which social outcomes they valued and then asked about the relationships among these opinions. Did each outcome provide a reason for a different one? Would each outcome cause a different one? If each outcome came to pass, would it make them more likely to support another outcome? Network diagrams derived from these responses represent portions of these individuals’ ideologies, understood as structures of political thought. Scrutiny of the network diagrams and analysis of the aggregate data suggest that most respondents carefully and reasonably identified relationships among their own ideas. Features of their networks predicted their assessments of five prominent politicians. This exploratory study paints a strikingly different picture of the sample than what would emerge from more conventional methods, such as factor analysis. Instead of a group that looks ideologically homogeneous on a unidimensional scale or that exhibits a low level of ideological coherence (because very few of their ideas are correlated), this method displays a collection of people who hold diverse and complex structures of thought. The method should be replicated with representative samples to explore the variation and significance of such structures.

See also: what if people’s political opinions are very heterogeneous?;  individuals in cultures: the concept of an idiodictuonideologies and complex systemsdon’t let the behavioral revolution make you fatalistic, etc.

seeking a religious congregation for a research study

I am seeking a congregation (of any religion, denomination, tradition, size, and location) for a research study. My interest is in testing a new method that I have been developing with colleagues that could apply to any community. I would give the congregation’s leadership–or its full membership–easy-to-understand findings about shared values and areas of disagreement within their congregation that should have practical value for planning events and programs.

Please consider whether this project might interest a congregation to which you belong or one that you know. Inquiries are welcome. More details follow:

I would ask the clergy or other leader(s) of the congregation to encourage members to take anonymous online surveys. The minimum would be two: a short survey with open-ended responses followed by a multiple-choice survey a week or two later that is based on the first one. I would be interested in repeating the multiple-choice survey months later to understand change, although that’s optional. If it’s practical, I would also like to visit and observe informally to get a feel for the community.

I would publish a scholarly study that would refer to the congregation anonymously (e.g., “a Protestant church in the Northeastern USA”). I would also provide the congregation with concise findings in PowerPoint format and would be happy to discuss them. No money would change hands. The congregation would own the PowerPoint and would not be obliged to publish or share it in any way. No individuals would be obligated to take the surveys, and I would expect only some people to do so. No identifiable information about individuals would be shared either within or beyond the congregation.

I could provide more detail about the method, but in brief, we don’t simply ask people their opinions about values, beliefs, and norms. Instead, we ask them how their personal opinions relate to each other. For instance, do they value A because they value B? Do they think that A causes B? From those responses, we generate network diagrams of the beliefs of each respondent and of the community as a whole. In this study, the questions would focus on religion and the congregation as a community, not on politics (unless respondents happen to bring up political matters).

Typically, each person’s responses are unique—a nice illustration of the uniqueness of human beings and how much we lose when we assign people to categories. Yet we typically see clusters of agreement and disagreement that can otherwise be overlooked. Understanding these patterns should provide ideas for visitors, readings, events, discussion groups (etc.) that would be valuable for the specific congregation.

what if people’s political opinions are very heterogeneous?

Here is what I will call a “top-down” theory of public opinion: Everyone should assess the most important issues of the day. An example might be the federal budget. Individuals are entitled to their personal beliefs or preferences about taxation, spending, and deficits; nevertheless, these three topics have an objective logical structure. If you want lower taxes and no deficits, you must favor spending cuts. If you want more spending, you must support either tax increases or bigger deficits. It is a civic obligation to think through your own complete view of these issues.

Analyzing a nationally representative survey of such matters should reveal that: a) most citizens hold opinions about them, and b) everyone recognizes the same underlying logic, even as they disagree about priorities. Citizens should vote according to their priorities, and the majority should prevail.

If, however, a) and b) are false, then democracy is flawed. Perhaps people should learn more and pay more attention, or the political system should offer greater clarity, or perhaps democracy is simply a false ideal.

Philip Converse’s “The Nature of Belief Systems in Mass Publics” (1964) has been enormously influential (cited more that 11,000 times). He does not employ exactly the framework with which I began this post, but his analysis suggests that democracy is a flawed ideal.

Analyzing surveys of political specialists, such as congressional candidates, Converse finds that these respondents meet both tests mentioned above. Professional politicians and other experts hold opinions on the great issues of the day that fit together in similar ways, even as they debate the contested questions. However, according to Converse, surveys of representative Americans reveal scant opinions and no consistent structure among opinions. Ordinary people are not, therefore, in a position to weigh in on the debates of the day.

Converse doesn’t use network analysis, but I can convert his core findings into network maps. (For those interested, I turned every tau-gamma coefficient above .05 in Converse’s Table 7 into an edge in these graphs.)

Congressional candidates share the following model of the political debate ca. 1962.

Whether candidates are Democrats or Republicans is strongly related to their views of federal interventions on education, housing, and employment, and those issues are linked together. Military aid to foreign countries is not connected–it stands alone as a topic. (“FEPC” means the proposed Fair Employment Practices Commission.)

Meanwhile, this is the result of using the same method with Converse’s data from representative adults at the same time:

There are no links at at all among the ideas (meaning no coefficients > .5). Converse suggests that people are not in a position to play the role expected in a “civics class” theory of democracy, assessing and voting on the issues of the day.

Now here is a rival, “bottom up” theory: The social world is enormously complex. Responsible people may hold opinions on myriad issues, and no one can have a view about everything. Even if we focus on a single domain, such as the federal budget, it is highly complex. Federal spending includes funds for many different purposes. Each federal program has a financial cost, but it also has moral significance, behavioral impact, symbolic meaning, likelihood of success, etc. If people are diverse and free, they will naturally form many different structures of opinions about the range of topics that they have considered so far.

Imagine a sample of people, each of whom holds a thoughtful and carefully considered view of public issues and connects each opinion to other opinions in rigorous ways. However, each person holds a unique structure. Then, if researchers contact each person and ask a short list of survey questions (out of the innumerable questions that they could ask), and the aggregate survey data is analyzed using the methods pioneered by Converse, it will reveal no significant links at all. The heterogeneity of individuals’ beliefs will yield very low coefficients at the group level. Their graph will look just like the second graph above–not because they are tuned out of politics or undisciplined in their thinking, but because they differ.

Converse (p. 44) explicitly acknowledges this possibility. He notes that if people’s opinions become “increasingly idiosyncratic” as they move away from the elite, we will find “little aggregative patterning of belief combinations” in their data, because they will not know “what goes with what” (presumably, according to the elites) and will “therefore put belief elements together in a great variety of ways.”

However, Converse counters with a different kind of evidence. According to his analysis of longitudinal data from the American National Election Studies from 1958-60, most people’s opinions on most issues change as if by sheer chance as time passed. This implies that they do not hold “well knit but highly idiosyncratic belief systems” (p. 47). Instead, they have no “meaningful beliefs” at all (p. 51). The exceptions are various small and specialized “issue publics” who hold stable opinions about specific concerns, including many Blacks and some White Southerners on questions of race.

On this point, I would only note that an enormous number of longitudinal studies have been conducted since 1962, and some find stability over time (e.g., Kustov, Laaker & Reller 2021), unexplained–and perhaps random–change (Jaeger, M. M. 2006), or a mix of the two, depending on this issue (Kiley & Vaisey 2020). For instance, according to the last of those studies, the proportion of individuals who change their opinions about abortion is extraordinarily low. Converse’s blanket claim that most people’s opinions on most issues change randomly seems overstated.

I do not assume that all of my fellow Americans (or anyone else) holds complex and rigorous mental models of politics. Some people hardly hold any opinions, or cannot connect their opinions together, or do so in illogical ways. My point is that we cannot tell from aggregate data what proportion of people hold thoughtful views unless we assume that all worthy views must manifest the same structure. Converse uses scare quotes around the word “proper” when he mentions “the sophisticated observer’s assumption of what beliefs go with other beliefs,” but if we seriously doubt that elites understand the “proper” arrangement, then we should expect regular people to exhibit great diversity.

In that case, we should seek to understand the heterogeneous structures of ordinary people’s views. This kind of research has three important purposes:

  1. It may counter the very low assessment of public capacity that is fashionable now in political science, deriving from Converse. That skepticism may prove self-reinforcing, since experts who doubt the public’s ability to reason will not be motivated to defend or improve democracy.
  2. It may yield specific findings about structures of thought shared by sub-groups. Those insights will be invisible as long as we only we ask whether the public shares the structures assumed by political elites.
  3. It offers a perspective on political institutions. Even if many people hold worthy but heterogeneous structures of ideas, we probably still need mass elections to address prominent topics, such as whether to raise or cut taxes. I support that kind of democracy. But it will always be frustrating, because the structure presented by elites will fail to match the reasonable structures in many people’s minds. Fortunately, we can complement mass majoritarian democracy with other forms of participation. Civil society provides an enormous range of specialized groups, and individuals can choose to join the groups that match their personal ideologies, exit ones that don’t, advocate for adjustments within the groups they belong to, and compete for new members. Meanwhile, relatively small and relatively discursive bodies, such as local town meetings or Participatory Budgeting sessions, offer individuals opportunities to share their idiosyncratic views in public. We might understand those venues as opportunities for deliberation that can yield wiser judgments. Or we might understand such venues as places where people can display their uniqueness, as Hannah Arendt advocated. Thus a “bottom up” theory supports decentralization, pluralism, or polycentricity.

See also: Moral and Political Discussion and Epistemic Networks; modeling a political discussion; individuals in cultures: the concept of an idiodictuon; ideologies and complex systems; don’t let the behavioral revolution make you fatalistic, polycentricity: the case for a (very) mixed economy, etc. I also acknowledge Jon Green, Nic Fishman, Sarah Shugars, and others who are working in this general domain today.

Moral and Political Discussion and Epistemic Networks

Yesterday, I got to give an International Society for Quantitative Ethnography (ISQE) Webinar on “Moral and Political Discussion and Epistemic Networks.” I really enjoyed the questions and conversation. This is the video of the whole event:

Abstract: An individual holds linked beliefs about political or moral issues, which we can model as a network. How these ideas are linked together influences the person’s actions and opinions. When individuals discuss, they share some portions of their respective networks of which they are conscious at the time. Some network structures are better than others for discussion: overly centralized or scattered networks are problematic. Individuals tend to demonstrate similar network structures on different issues. Thus, relying on certain kinds of networks is a character trait. People, with their respective networks of ideas, are also embedded in social networks. An idea is more likely to spread depending on features of both the social network and the idea networks of the people who interact. As a whole, a population may develop a shared network structure. An idea that is widely shared and frequently central in individuals’ networks becomes a norm. Institutions are partly composed of such norms. A community or a culture is a single network with disagreement. Ultimately, all such networks interconnect. This theory has implications for politics, ethics, and research methodologies.

Additional ISQE webinars are listed here.

modeling a political discussion

In 2015, students at my university, Tufts, and at Kansas State University discussed the same topic–the social determinants of health–in online forums. Colleagues and I analyzed the text in a novel way. The underlying theory is that conversations can be modeled as networks, where the nodes are specific ideas and the links are reasons and other connections that people assert. Specifically, we sought to use Epistemic Network Analysis (ENA) to model and compare the conversations. ENA is now very easy to use online and I have been playing with it for other purposes. One of its distinctive features is its ability to locate specific ideas at meaningful locations on a two-dimensional graph so that you can see dimensions of agreement and disagreement.

Our results are now published as Peter Levine, Brendan Eagan & David Williamson Shaffer, “Deliberation as an Epistemic Network: A Method for Analyzing Discussion,” in Barbara Wasson and Szilvia Zörgo (eds.), Advances in Quantitative Ethnography, proceedings of the Third International Conference, ICQE 2021 Virtual Event, November 6–11, 2021 (Springer Switzerland, 2022), pp. 17-33.

The image I reproduce with this post shows one particular ENA visualization of the discussions. The one at Tufts is in blue; the one at KSU is in red; and the one in the middle shows the difference between them (literally, the Tufts network subtracted from the KSU network).

You would have to read our paper to get a full explanation, but here is a glimpse. Basically, the Tufts students tended to connect inequities in health with race and class. Some of the KSU students also made those connections, but some of them drew connections between bad health outcomes and personal behavior, for which they blamed individuals’ upbringings. Thus the KSU discussion roughly looks like a triangle with three corners (race and class, personal choices, and health outcomes), while the Tufts discussion omits one of those corners.

The same result might have been clear enough from a conventional approach–reading, interpreting, and (possibly) coding the transcripts. However, we argue that the plausibility of the ENA findings validates the method, which can then be used to model other discussions.