Computational and simulation modelling research on political ideologies has evolved from measurement concerns to cognitive and complexity‐oriented approaches. While the former were rooted in the debate on stability of political attitudes and beliefs in mass electorates as initially formulated by Converse (1964) and further argued by Achen (1975) and Zaller (1992), the latter were rooted in the debates on the role of ideology in social and political change. It is this latter orientation of research which interests our presentation of computational modelling of ideologies. Though it is based on political attitudes, it has induced two types of complexity‐based modelling approaches: (a) those targeting phenomena of social structure emergence and (b) those targeting phenomena of political order emergence; combinations of these two approaches are often targeted. The modelling approaches in the former class of phenomena have been generated by the definitional and theoretical approaches of ideologies as complex combinations of attitudes toward social and political major issues. The models in the latter class have been generated by the dynamic modelling approaches on ideology change.
This part presents some of the most relevant ideology computational and simulation modelling approaches which are based on political attitudes. Ideology computational modelling employs methods for approaching complex phenomena like emergence of structure and/or order, and self‐organization processes like clustering of political attitudes and polarization of ideology.
Ideology has been defined and modelled as a cluster of attitudes (Campbell et al., 1960), belief systems (Converse, 1964: 2–3), systems of attitudes viewed as structures organized on multiple dimensions (McGuire, 1989: 49) and organized attitudes, beliefs and values (Rokeach, 1968, 1973). These conceptual backgrounds have induced a development of ideology modelling approaches based on structures of political attitudes and beliefs and their variation in electoral communication campaigns. This type of modelling approach has often assumed a political culture perspective over ideology.
On the other hand, ideology has been defined and modelled as belief systems about social and political order (Parsons, 1951; Campbell et al., 1960; Kerlinger, 1984; Erikson and Tedin, 2003; Jost et al., 2009a), institutional aspects included (Denzau and North, 1994). In their comprehensive review of literature on the ideology theory and ideology formation and change modelling, Jost et al. (2009a) provided a synthesis of the traditional definitional and conceptual frameworks as well as the new trends emphasized by the qualitative models in this area. Their approach is relevant from a theoretical perspective, which succeeds in covering the major problems and orientations in the ideology theory on political, psychological and social science backgrounds. The major developments they identified at the theoretical level are relevant for modelling theory as well, explaining the main type of approaches in this respect.
A similar approach, but from a different perspective, was elaborated by Homer‐Dixon et al. (2013), who made a synthesis of the modelling paradigms and trends in ideology research which cover more or less the same time interval. Their approach is mainly methodological and emphasizes the modelling approaches which have employed computational and simulation technologies in their conceptual and methodological modelling designs. They identify a spread of focus on multiple, though narrow, aspects of ideology modelling based on computational and simulation technologies, thus providing arguments toward their own strategy of developing an integrative modelling approach on ideology. As computational and simulation technologies have advanced incredibly fast during the past three to four decades, the conclusion they draw regarding the cleavages induced by these technological developments fully justify their approach. Two reasons may have induced a trend toward the specialization in ‘niche’ issues within modelling theory and methodology research:
By employing different concepts and basic assumptions, the definitional approaches to ideology have induced in the ideology modelling research choices for various types of paradigms revealing the social, cognitive or psychological dimensions and perspectives of analysis. Schematically, a brief dichotomy‐based description would best describe the ideology modelling typologies from a paradigmatic point of view. Ideology modelling research has been founded on the classic conceptual dichotomies which have marked its various definitional approaches: left versus right, individual versus group and cognitive versus social.
The left–right model of ideology has inspired not only the spatial modelling approaches, but also various other models which view ideology as combining social, cognitive and psychological dimensions as well as economic and political conservatism: Jost (2009: 134–135) elaborates a motivated social cognition model of ideology by taking into analysis relevant dichotomies, like the opposition between the need for stability and the desire for change (Jost, 2009: 129).
The spatial modelling paradigm has been employed in the development of ideology formation and change models, like the unidimensional left–right spatial models (Enelow and Hinich, 1984, 1994; Enelow, Hinich and Mendell, 1986; Hinich and Munger, 1996). In this type of paradigmatic approach, typically developed within the rational choice conceptual framework, the notion of utility is employed so as to define a construct of ideology in which socio‐economic and socio‐political principles combine in providing an explanation of human action choice as determined and/or accompanied by ideology arguments. A second type of spatial modelling paradigm provided for bidimensional models, which have typically involved value theory (Braithwaite, 1997), psychological and sociological theories on political attitudes (Eysenck, [1954]1999) and political culture theories (Inglehart, 1997). A third class of spatial modelling paradigms has provided for the multidimensional models inspired from culture theories (Geertz, 1964). Multidimensional spatial models have emphasized a strong holistic character due to their approach on ideology as a phenomenon involving multiple topics and multidimensional attribute or cognitive spaces. Tendencies toward holistic modelling approaches which take into consideration the ‘whole’ of the factors which provide for ideology formation and change and for its measurement were claimed during the early 1970s (Barnett, 1974; Serota et al., 1975, 1976; Barnett et al., 1976). Later on, the tendency toward integrative models claimed by Homer‐Dixon et al. (2013) developed towards cognitive and complexity‐based paradigms.
Another classic dichotomy employed in ideology modelling is individual versus group. Jost and collaborators (Jost et al., 2009a,b) classified ideology conceptual modelling approaches into top‐down (specific to political science approach) and bottom‐up (specific to psychological approach).
The dual‐process model of ideologies (Duckitt, 2001) is based on political attitudes and their organization on single or multiple cognitive dimensions of preference and judgment (Jost et al., 2009a). The model starts from the hypothesis that attitudes toward social and economic issues are distinct and they could predict ideological patterns of preferences. Individuals’ social dominance orientations are based on a power‐struggle view of the world and predict economic conservatism as prevalent in comparison with social conservatism, while individuals’ right‐wing authoritarianism orientations are based on a view of the world as threatening and predict social conservatism as prevalent with respect to economic conservatism. The individual orientations provide indicators to the functional structure of ideologies so that they are analysed as good predictors of the intergroup ideological positions (Duckitt et al., 2002; Duckitt and Sibley, 2010). The dual‐process model thus combines top‐down with bottom‐up modelling. As a methodological approach, dual‐process modelling has so far employed structural equations – see Duckitt et al. (2002) and Weber and Federico (2007).
The theoretical models of ideology have inspired or have been based on numerous empirical modelling approaches, many of them addressing the political attitude issues. Notwithstanding conceptual richness, theoretical broadness and effervescence of search for empirical support, very few political attitude‐based computational models have been developed on these quite impressive backgrounds. The proper explanation to this situation resides in the need for an appropriate computational modelling technology able to cope with the complexity of the ideology issue, as well as in the need for an integrative paradigmatic approach able to overcome the cleavage‐prone tendency and also the tendency to spread the focus in ideology theoretical and empirical modelling.
This part selects some of the ideology computational and simulation models which are relevant as political attitude‐based models. Regardless of the historical age to which they belong, each of these models represented a breakthrough in its time for identifying a proper methodological and technological modelling solution for an important theoretical problem. Other computational models of ideology, which are not necessarily based on political attitudes, have been developed in areas like electoral studies, opinion dynamics, voting behaviour, political parties and partisanship. Though not included in the area covered by this book, and also for reasons concerning the space available in this part, some of these latter models are mentioned briefly as they are more or less connected with the issue of ideology modelling.
The classic definition of political ideology which provided initial support to ideology computational modelling is based on political attitudes toward objects like parties, candidates and policies.
During the 1950s, the Columbia model introduced a social influence paradigm in modelling ideology change. The Columbia studies, elaborated by Paul Lazarsfeld and collaborators (Lazarsfeld et al., [1944]1968, 1948; Berelson et al., [1954]1986), focused on the social context of political (voting) preference formation, taking into consideration variables describing social status, residence and religious preferences. The Michigan model developed by Angus Campbell and his team (Campbell et al., 1960) approached ideology formation from the point of view of the interest in political participation (voting), policy issues (governmental action) and political loyalties (party identification).
The 1960s were dominated by the political culture approach to ideology. Converse’s is an empirical model based on the study of several variables used as the predictors of ideology orientation of an individual voter. In the same period, Gabriel Almond and Sidney Verba (1963) introduced a political culture theory of ideology formation.
Converse defined the concept of ideology as an organized structure of attitudes and beliefs, and provided the first conceptualization (constraint theory) and operationalization (correlation) of a belief system related to political attitudes and values. The constraint theory has been the subject of debate, as Converse’s model has been criticized for his unidimensional approach based on the consistency between attitude (issue position) and ideas which make up a belief system.
Campbell et al. (1960: 189–190) approached ideology as a ‘cluster’, in which attitudes are organized on a functional criterion illustrating a ‘means‐and‐ends’ view (Campbell et al., 1960: 189). The functional approach has further served for identifying a class of models which are based on ideology as a response to individual needs. As such, this suggests an opposed perspective to the models based on institutions or power modelling. The political culture modelling approach to ideology, as a modelling tradition initiated by Converse (1964) and Almond and Verba (1963), focuses on political attitudes and belief structures, defining their functionalities on a rationality basis.
These models represent the theoretical and empirical foundation and inspired ideology computational modelling approaches developed a decade later, as soon as the computer technologies offered solutions and computer expertise had diffused among sociologists and political scientists.
Two main modelling paradigms are addressed in the following section: spatial and cognitive. Developed in the 1970s, the political information processing theories as well as the cognitive theories in ideology research inspired ideology cognitive modelling, which includes political reasoning, political judgment as well as political cognition issues. It is only lately that the integrative spirit has gained ground and proved more powerful as the technologies of ALife have provided ways to unite these and other computational paradigmatic approaches into complex, more believable ones.
In the early 1970s, scale‐based attitude measurement was facing a challenge which made it appear a rather poor and slight means to measure the complex structures of political attitudes and beliefs viewed as ideology patterns. An attitude scale was meant to organize individual self‐reporting ideological positions of the respondents so that preferences and their intensity could be measured in a population. However, extending the attitude scale‐based measurement to ideology might have proved a very difficult task as political ideology is defined as a structure of political attitudes and beliefs. Ideology measurement would thus have to approach complex ideological patterns that include attitudes, beliefs and values altogether as structural attributes.
Holistic conceptualizations of ideology measurement became salient as value theory research provided a different approach on the issue of ideology and its structure by defining it as organized structures of attitudes and beliefs (Rokeach, 1968). Though this conceptual perspective has allowed for substantial advances in the study of ideology as more complex than a multicriteria evaluation of attitudes toward objects with ideological significance, it has nevertheless revealed a major weakness of attitude scale‐based measurement: data provided by attitude surveys do not reveal whether the selected attitudes are relevant enough to correctly identify the attributes and the structure of an ideology (Serota et al., 1976). In order to obtain the set of relationships among relevant attributes of an ideology, an entire population should be investigated with respect to a predefined list of concepts which are supposed to characterize an ideology as a whole.
A first attempt to solve this requirement aimed at providing a holistic empirical modelling approach on ideology was that of considering the ideology as a culture pattern: Scott’s (1959) idea was to describe ideology as a cluster of interrelated culture items which could be described on a correlation basis, such that, for several groups, the differences between items revealed the ideological difference between groups. Geertz’s (1964) idea was to define an ideology as culture and to view it as a spatial representation (i.e. a map) such that each object (i.e. culture item) would have a location and a relative distance to other objects on the map.
Serota et al. (1976) follow both ideas and define an ideology as a culture with a multidimensional spatial representation achieved with the multidimensional scaling method. This approach measures an ideology by evaluating discrepancies between concepts as distances in the spatial representation.
This type of model defines a political ideology as a structure of culture items following a spatial representation inspired by a spatial analogy with a ‘map’, an idea introduced in the earlier modelling approach developed by Geertz (1964).
Ideology is modelled as a set of concepts. By means of the multidimensional scaling technique, the concepts are transformed into points represented in a multidimensional space such that each point could be projected on each dimension (axis) of this space. This representation is used for defining and evaluating the differences between objects in this space as distances. The notion of ‘distance’ is employed as a means of operationalizing the judgments concerning the dissimilarities (discrepancies) between concepts. Evaluating the dissimilarity (discrepancy) between any two concepts allows for the construction of a matrix of distances between concepts. The multidimensional scaling method allows the transformation of this matrix into a spatial configuration of the concepts and their relationship in the ideology structure.
The experimental procedure combines two steps:
Ideology structure is achieved as a configuration of objects (relevant concepts) in a multidimensional (cognitive) space. The model is based on the method of multidimensional scaling used in spatial modelling of political attitude change. The model has been used (Serota et al., 1976) in measuring the ideological differences between two specific groups defined as ‘upper class’ and ‘lower class’. The model has also been employed in measuring the effects of interpersonal and media communication on ideology in empirical longitudinal studies on several groups of individuals (Barnett et al., 1976). Measurements of the communication impact on the political attitudes and ideology change in several groups have been reported with respect to the 1974 US congressional elections. The measurement of discrepancies between concepts as an effect of interpersonal/media communications revealed homophily‐induced modifications of the distances between concepts.
The evaluation of the impact of the concepts in the political messages (stimuli) on the ideology structure has been used in the design of message strategies during the electoral campaign: messages which transmit information about a relationship between two concepts would result in making the concepts ‘move’ in the space to closer positions (i.e. their distance on given dimensions in the multidimensional space will diminish). The model predicts group behaviour with respect to political candidates and parties by measuring the discrepancies between groups’ political attitudes and parties’ or candidates’ ideology.
Classic political science modelling approaches, starting with the Yale model elaborated during the 1950s until lately, have been focused more on the mechanisms and processes of attitude formation and change which are based on media and political communication during electoral campaigns and also by the communication and political persuasion in the networks of interpersonal relations. Starting in the 1980s, this latter topic was intensively investigated with regard to the polity macro‐level effects of the individual interactions at the micro level. The complexity of the political attitude phenomena has often represented a difficult aspect in modelling research approaches. The various mechanisms of interaction between individual agents as voters need further investigation with respect to the processes they control and the emergent phenomena they generate. Designs of models have only lately involved complexity‐based studies of the systemic effects of ideology‐based interactions among individual agents in networks of interpersonal relations.
The list of political ideology models thus includes various topics and types of approaches, like political discourse (Leifeld, 2014), language and thought (Dirven et al., 2003), ideology formation and ideological polarization (Baldassarri and Bearman, 2007; Lorenz, 2014), dynamic parties and coalitions (Fowler and Smirnov, 2005), political insurgence, uprisings and revolutions (Lang and De Sterck, 2012), oligarchs, ideologic preferences and voting behaviour (Wright and Sengupta, 2015), political belief systems (Homer‐Dixon et al., 2013) and ideology diffusion in social networks (Brzozowski et al., 2008), to name but a few topics.