Mapping the moral domain in a modern town hall: Moral foundations’ effects on collective action conditioned by ideology and issue

Abstract

We examine a year of political debate in a modern town hall to understand the following: when does moral talk turn to talk about action? For which ideological groupings do specific moral foundations matter and in the context of which issues? We analysed 210,170 tweets about the Australian television programme Q&A – a panel show with synchronous live and online audiences. We examined the topics, prevalence of different moralities and how these moralities explain talk of collective action for different ideological groups. Across all issues and ideological groups, talk of action was most often associated with talk of authority but ideological differences between left- and right-leaning participants emerged. Among left-leaning people, action was motivated by a range of moral foundations, but among right-leaning people action was primarily motivated by authority. Our findings map the ideological diversity found within modern town halls and confirm the critical role of these settings in contemporary socio-political life.

Keywords

Collective action modern town hall moral foundations natural language processing political ideology

Throughout history, town hall meetings have played a critical role in the regulation of social and political life within communities. Within such meetings, authorities and citizens discuss, deliberate and plan actions for the betterment of their community and society – reflections of an ‘iconic form of direct citizen democracy’ (Field, 2019: 2). With the advent of globalisation and the development of new media, town hall meetings changed (e.g. Forsyth, 2013; Lukensmeyer and Brigham, 2002). Indeed, Field (2019) laments the decline of town hall meetings and links them with a precipitous decline in democratic norms in general. This article is concerned with a mixed media approach to understand the modern town hall, exemplified in the Australian television programme Q&A. The programme invites panellists, such as public figures, politicians and experts, to engage in live discussions with the Australian public on various social and political issues (ABC, 2023) – similar variants exist in other countries (e.g. Question Time on the BBC). In addition to live discussions, Q&A encourages the broader public to engage in the conversation through synchronous and asynchronous online interactions on the social media platform Twitter/X. In this way, the traditional town hall has arguably progressed to encompass multiple modes of media (television, social media interactions) and includes those who are not physically co-present.

Town halls have historically been a primary way in which authorities are held to account, with deliberation and debate commonly arising from people having different views about social and political life. We aim to understand the diversity that exists within such real-world manifestations of political interaction. Given that town hall discussions revolve around notions of what is ‘right’ versus ‘wrong’, examining people’s moral worldviews may help to understand this diversity. Importantly, a typical outcome of town hall meetings is to address what should change or what action should be taken within the community (Field, 2019; Robinson, 2011). Therefore, we are especially interested in the relationship between moral talk and talk of taking collective action to change the status quo (see Thomas et al., 2022a).

It is also the case that town halls range in terms of the political ideologies of those who are present. Most communities are pluralistic with different views about the ‘ideal arrangement of society’ (i.e. ideology; Cohrs, 2012: 55). We look at how moral talk may differ among different ideological groupings, captured via the distinction between centrist, left- and right-leaning people. Finally, town hall meetings range in the issues that are canvassed, especially in an open democratic forum such as the Q&A programme. Given that different moral worldviews may be implicated in some issues more than others (e.g. Feinberg and Willer, 2013; Reimer et al., 2022), we focus on whether moral talk may explain action in some contexts (and for some ideological groups) but not others. Accordingly, bringing these points together, we examine the relationships between (different forms of) moral language use and talk of taking action with others to bring about a desired change. We look at these relationships in the context of specific ideological groupings and across different issues. Figure 1 provides a conceptual overview of our approach.

Figure 1.

Conceptual figure.

Does moral talk turn to talk of action?

Town halls are where people discuss what is right and ideal, wrong and undesirable – in this research, we capture these in terms of theories of moral reasoning (see Table 1). When a person holds a moral worldview about what is right versus wrong, this is known as endorsing a moral value (Skitka, 2010). Moralising an issue is an inherently interactive process whereby an individual engages with or is persuaded by new information about the moral significance of an issue (Mooijman et al., 2018). When an attitude is held with moral conviction (i.e. it has become moralised; Skitka et al., 2005), this has several consequences. First, individuals tend to hold a strong and absolute/universal stance on that topic (Ryan, 2017; Skitka et al., 2021). Critically, when moral values are in dispute, people with a moralised attitude are more likely to support actions to defend or protect their beliefs (Milesi and Alberici, 2018). For example, holding an issue with a strong, moralised stance predicts political action, including voting, activism and collective action (Agostini and van Zomeren, 2021; Milesi and Alberici, 2018) as well as more radical actions (Mooijman et al., 2018; Pauls et al., 2022). Collective action refers to actions that people perform on behalf of a group to promote that group’s goals (Wright et al., 1990). In the context of online interaction, constructive discussions can validate and strengthen moral judgements about an issue, in turn increasing support for collective action towards a political cause (Alberici and Milesi, 2018; Smith et al., 2018). Thus, endorsing strong moral values may influence support for collective action to bring about a desired change (see Figure 1).

Table 1.

Definition and example words for the five moral foundations.

Moral foundation	Definition	Example words from Extended Moral Foundations Dictionary
Care/harm	A desire to protect and care for others’ well-being	Compassion (+), cruel (–), distress (–)
Fairness/cheating	A desire to protect others’ rights and sense of justice	Equality (+), fair (+), discriminate (–)
Loyalty/betrayal	A feeling of loyalty towards one’s group including being prepared to self-sacrifice for the group	Supporter (+), solidarity (+), stealing (–)
Authority/subversion	A valuing of leadership and respect for authority and tradition	Control (+), compliant (+), disrespect (–)
Sanctity/degradation	A desire to live a noble life and maintain purity	Noble (+), God (+), disgusting (–)

‘+’ indicates the word is positively valanced for that moral foundation, whereas ‘–’ indicates the word is negatively valanced for that moral foundation.

Prior research has established that morality is pluralistic. That is, different people (or groups) can be guided by different moral values, and by more than one value at a time (Milesi and Alberici, 2018). Moral Foundations Theory (MFT; Graham et al., 2011) captures five different cross-cultural moral convictions or foundations which people can be motivated by. These foundations include care/harm, fairness/cheating, loyalty/betrayal, authority/subversion and sanctity/degradation (see Table 1). Importantly, the foundations themselves reflect broader composites of attitudes, cognitions, emotions and linguistic styles (Frimer, 2020). For example, someone who endorses fairness may support universal human rights and be more likely to talk about equality and justice. As such, language can be a valuable window for examining these foundations, and discourse can give insight into the moralising process as something that is interactive and didactic (Boyd et al., 2022; Hoover et al., 2020). While prior research has examined the relationships between moral foundations and socio-political attitudes (see meta-analysis by Kivikangas et al., 2021), the role of moralising language use in social interaction has been relatively less studied.

Moreover, moral foundations have been linked with specific forms of action. For instance, people who endorse sanctity were less likely to vaccinate against COVID-19, while those who endorse fairness and group loyalty were more likely to do so (Reimer et al., 2022). As such, the moral foundations people endorse can influence their willingness to engage in action to bring about (or prevent) a desired change (see Figure 1). Given that town halls are places where people advocate for action to bring about change, we ask, ‘When does moral talk relate to talk about action?’

Does moral talk differ based on political ideology?

A further complexity is that according to MFT, the moral frameworks which people are motivated by can differ as a function of their political ideology (see Figure 1; Graham et al., 2011). There is evidence that liberals (left-leaning) rely mostly on care- and fairness-based moral frameworks to navigate social and political issues (Graham et al., 2009; Haidt and Graham, 2007; Kivikangas et al., 2021). In contrast, moral foundations around loyalty, authority and sanctity are more commonly associated with a conservative (right-leaning) political orientation (Graham et al., 2009; Haidt and Graham, 2007; Kivikangas et al., 2021).

The Intergroup Value Protection Model (van Zomeren et al., 2023) posits that people are motivated to protect the moral values which underpin their group memberships, and they do so by holding a moral conviction, protesting and/or supporting other forms of collective action. Indeed, Thomas et al. (2022b) found that for (right-leaning) supporters of a right-wing rally in Australia, their online engagement and support for the rally was shaped by moral values of care, fairness, loyalty and authority. In comparison, (left-leaning) opponents endorsed loyalty and appeals to authorities to uphold their opposition to the rally (Thomas et al., 2022b). Curiously, the right-leaning supporters relied on moral frameworks which are more commonly associated with left-leaning people, and vice versa. Other work also shows that differences between liberals and conservatives’ moral rhetoric in political interactions are not always consistent with traditional readings of MFT (Garten et al., 2016; Roy and Goldwasser, 2021). These findings demonstrate how the same word or construct may have different meanings depending on the (social, political) context in which it is deployed (Boyd and Schwartz, 2021). Indeed, it is recognised within the moral psychology literature that MFT does not apply unconditionally, as context plays a role in the foundations endorsed by different political ideologies (Haidt and Joseph, 2011; Suhler and Churchland, 2011). Thus, positions about right and wrong within town hall discussions may differ not only by ideology but also by the issue discussed (see Figure 1).

Moral talk in context: does the issue change the effect of moral foundations and political ideology on collective action?

Some moral foundations are more relevant to one issue than another (see Figure 1; Haidt and Joseph, 2011; Suhler and Churchland, 2011). For instance, a moral value of care/harm is relevant to climate change support (Feinberg and Willer, 2013), while respect for authority and preserving hierarchies explains variation in anti-immigration interactions (Grover et al., 2019). Moreover, Richardson and Conway (2022) demonstrate that the nature of a movement itself can influence the relationship between moral foundations and collective action: values of care and fairness explained people protesting in support of Black Lives Matter, while values of authority, loyalty and sanctity were more relevant to people who supported a counter-movement (i.e. Blue Lives Matter). Yet, relatively few studies have considered the effects of morality across multiple issues within a single setting. Adopting a macro, descriptive approach to model many elements side-by-side (moral foundations, action, ideology, issue) may be particularly important for understanding the complex, real-world manifestation of political interaction in town halls. We examine the interplay between political ideology, moral foundations and collective action, in the context of three different issues covered in a years’ worth of political debate in an Australian town hall: Australian society and identity, climate change and the environment, and COVID-19 (see Figure 1).

The current research

With technological change, people now primarily engage with socio-political debates through multiple modes (live discussions, television, social media interactions). We conceptualise these as modern town hall discussions and analyse the Q&A programme as a distinct context for mapping a year of socio-political interaction in Australia. Specifically, the Q&A dataset provides a snapshot of Australian political discussion in a year of considerable socio-political upheaval. The year 2020 was marked by devastating wildfires across Australia and rising speculation about climate change, along with a global pandemic (COVID-19). In the context of border closures and a changing global order were discussions about the Australian identity and society more generally. We focus on examining how the general public moralises, and talks about acting upon, three different social and political issues (Australian society and identity, climate change and the environment, and COVID-19). We ask, ‘When does moral talk turn to talk about action?’ and ‘For whom (which ideological groupings) do specific moral foundations matter and in the context of which issues?’

To address these questions, we first characterise the online interactions for each issue by taking a data-driven approach using Structural Topic Modelling (STM). We then focus on the prevalence of moral content in these interactions by quantifying the degree to which different forms of moral language arose within each issue, for different ideological groups (see Figure 1). Finally, we explain how particular forms of moral language uniquely relate to talk of taking collective action on that particular issue by combining these more descriptive (inductive) approaches with a deductive approach using hurdle models.

We draw on natural language processing of Q&A Twitter/X data to detect patterns of language use reflecting moral foundations and collective action. We use the moral foundations (Hopp et al., 2021) and collective action (Smith et al., 2018) plug-ins to the Linguistic Inquiry Word Count tool (LIWC; Boyd et al., 2022), which identify the percentage of words or word stems representative of the pre-defined dictionaries. LIWC has been extensively adopted in social science research (Boyd and Schwartz, 2021; Tausczik and Pennebaker, 2010), and the moral foundations and collective action dictionaries have been validated to show that language-based psychological inferences credibly reflect underlying (cognitive, emotional) psychological states (Hopp et al., 2021; Smith et al., 2018). This ‘bag of words’ approach has the benefit of being accessible to people without coding or data science expertise; it is transparent (i.e. the dictionaries for the categories are in the public domain) and a standardised dictionary enables ready comparison across different groups and contexts (Boyd and Schwartz, 2021; Tausczik and Pennebaker, 2010).

One statistical complexity with using LIWC is that the processed data are often zero-inflated and non-normally distributed (Bhattacharya et al., 2014; Xu and Zhang, 2018). That is, there are instances where the language of interest is not present, and a zero is assigned. To manage this aspect of the data, we employ hurdle models (Hu et al., 2011; Mullahy, 1986). Hurdle models are two-part count data models which, first, indicate whether an outcome occurred (as a binary variable) and, second, (after the hurdle is ‘crossed’) estimate how much the outcome occurred (as a continuous variable; Hu et al., 2011). In the current study, we distinguish between whether moral talk was associated with talk of collective action (i.e. as a binary variable; action language present or absent), and if so, by how much (i.e. as a continuous variable; degree of action language).

Finally, given the heterogeneity in the community and that political ideology can shape what kinds of moral foundations matter for predicting action, our approach examined how moral- and action-based language may be similar/different for those who approach the debate from different ideological backgrounds. We used an ideology classifier (Ram et al., 2025) to categorise people’s political ideology as left, neutral, right or far-right.

Method

Data collection

Social media data were collected from Twitter/X’s standard live stream on discussions related to the Australian television programme Q&A in 2020. The filter keyword ‘qanda’ was used to collect a total of 754,475 tweets spanning all 44 episodes of Q&A which screened from February to November 2020 (see Supplemental Table S1 for a summary of episodes). We used a subset of these tweets: a corpus of 210,170.

Research approach

Our research approach involved five steps of analysis. We first identified the political ideology of Twitter/X users via an ideology classifier (see Measures; Ram et al., 2025). Next, we used STM to characterise the online interactions. In a third step, we conducted natural language processing using LIWC-22 (Boyd et al., 2022) to classify tweets in terms of their attributes of moral language and collective action language. Using this output, the fourth step examined the overall presence of moral language and whether the language was generally positive or negative in sentiment. As a final step, we conducted a series of hurdle models to examine whether moral language was associated with collective action language across ideological group and issue. Language use reflecting the five moral foundations (care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, sanctity/degradation) was treated as a continuous predictor variable of collective action language (present or absent, degree of action language). There were two categorical moderators: political ideology (left, neutral, right and far-right) and issue (Australian society and identity, climate change and the environment, and COVID-19).

To assess the validity of the computerised LIWC scoring, we hand-coded a stratified sample of N = 360 tweets (n = 30 for each of the three issues by the four ideologies) and compared human codes to the moral foundations and collective action dictionaries. The agreement for moral foundation scores (88.1%), moral foundation sentiment scores (81.5%) and collective action scores (94.5%) was acceptable (see Supplemental Table S2 for the agreement and precision/recall for each category). The full validation methodology is reported in the Supplementary Materials.

Measures

Moral foundations

The Extended Moral Foundations Dictionary (eMFD; Hopp et al., 2021) was used to identify language reflecting care/harm, fairness/cheating, loyalty/betrayal, authority/subversion and sanctity/degradation (see Table 1). The dictionary provides two kinds of output: sentiment values indicate the valence of the language (e.g. positive vs negative values indicate care- vs. harm-based language, respectively) and score values represent the overall prevalence of the language per moral foundation (i.e. higher scores indicate higher prevalence; Hopp et al., 2021). We use the sentiment values to first provide a descriptive overview of the language use across the five foundations; then, we use the score values in the hurdle models to look at associations between moral and collective action language.

Collective action

Language use reflecting collective action (e.g. act, protest, march) was measured by the Collective Action Dictionary (Smith et al., 2018). Higher scores indicate a higher prevalence of action language.

Political ideology

We utilised an ideology classifier developed by Ram et al. (2025) to identify the political ideology of Twitter/X users. Ideology was classified based on the media sources that each user shares and the average political leaning of those sources. The Left-Right Media Publication Proxy classified users as left (195,656 tweets; 93.1%), neutral (8602 tweets; 4.1%) or right (1330 tweets; 0.2%), and the Media Bias Fact Check Media Publication Proxy classified far-right users (5582 tweets; 2.7%). The proxies were trained on five social media datasets (three Australian and two American) with 1.2 million profiles.

Issue

There were 44 televised episodes of Q&A in 2020, and each episode covered multiple issues. We manually coded the episodes into three discrete issues: Australian society and identity, climate change and the environment, and COVID-19 (see Supplemental Table S3 in the Supplementary file for a detailed breakdown). The number of tweets per issue was 68,989 for Australian society and identity, 75,506 for climate change and the environment and 65,675 for COVID-19.

Results

Describing the content: topic models

To assess the validity of the dataset based on interactions about Australian society and identity, climate change and the environment, and COVID-19, we first conducted unsupervised STM on unigrams using the stm package in R v4.3.2 (Roberts et al., 2014). STM assumes a given text has clusters of semantically relevant ideas (‘topics’) which can be identified by capturing correlations between word frequencies. All code for analyses is available on the Open Science Framework (OSF) at: https://osf.io/5g7dv/overview?view_only=1c85c25f68ac4c1fbd969960c675cb16. For each issue, tweets were pre-processed using the tm package to remove whitespace, numbers, punctuation, weblinks, English stop-words; convert words to lowercase; and lemmatize words. Special abbreviations specific to the dataset were also adjusted (e.g. auspol = australian politics). We removed the keyword used to collect the data (‘qanda’), and case-specific stop words (e.g. abc, abctv). Finally, we removed sparse terms with frequency counts of <2% or >99%. The final number of tweets per issue was 63,658 for Australian society and identity, 71,819 for climate change and the environment and 62,471 for COVID-19.

To decide on the optimal number of topics for each issue, we used a combination of fit statistics and visual inspection of the models (Brookes and McEnery, 2019; Kosinski et al., 2016; Roberts et al., 2014). After running STM models with k ranging from 5 to 25, model fit was assessed using the trade-off between semantic coherence and exclusivity, with the optimal solution chosen to maximise coherence while retaining sufficient exclusivity (Mimno et al., 2011; Roberts et al., 2014). Figure 2 shows that optimal k ranged between 10 and 15 for both Australian society and climate change, and between 13 and 17 for COVID-19.

Figure 2.

Statistical fit of models with different k.

We then visually inspected this narrower range by examining high probability terms, high FREX terms (a weighted measure of word frequency and exclusivity), Rank-1 values (Evans, 2014) and example tweets for every k. We identified a final model of k = 12 for Australian society, k = 12 for climate change and k = 15 for COVID-19. In the interests of space, we present the three most prevalent topics (per issue) to provide a broad overview of the content of each issue. Figure 3 displays word clouds of the most frequent words assigned to each topic (per issue).

Figure 3.

Word clouds for the top three frequent topics for each issue.

As shown in Figure 3a, conversations about Australian society and identity included the equal representation of social groups in socio-political discussions (topic 1); domestic violence, especially in Indigenous communities (topic 8); and political figures’ ability to handle crises (topic 10). Interactions about climate change and the environment (see Figure 3b) focused on holding Australian politicians accountable for climate action (topic 2), identifying inconsistencies between scientific evidence and political action (topic 3), and called for politicians to take responsibility for the Australian wildfires (topic 11). Finally, discussions about COVID-19 (see Figure 3c) demonstrated a sense of solidarity with, and gratitude for, a specific political figure (i.e. Victorian Premier Dan Andrews; topic 6) while also covering COVID-19 prevention (topic 12) and dissatisfaction with government leadership (topic 13). See Supplementary Materials for a more detailed summary of the topic model results. Our analysis below draws, inter alia, on this content in interpreting key findings.

Quantifying the moral content in online interactions

Next, we examined the prevalence of moral content in the online interactions about the different issues, for the different ideological groups (see Figure 4). The sentiment scores from the eMFD indicate the valence of the moral language and were thus well suited to providing a sense of whether a particular foundation was discussed in positive (e.g. care) or negative (e.g. harm) terms, and to what degree. Figure 4 shows that, across all four ideological groups, there was similar use of negatively valanced language which focused more on harm (vs care) and degradation (vs sanctity) when talking about all three issues. In general, the different ideological groups were unexpectedly similar in as much as all used more negatively valanced moral language. However, the moral language of the ideologically neutral group was more positively valanced, suggesting that politically partisan groups may be more critically engaged in socio-political conversations.

Figure 4.

Heat map of sentiment of moral language across ideological group and issue.

However, one distinction between left- and right-leaning people was their use of loyalty-based language. Left-leaning people spoke more about loyalty, whereas ideologically right people spoke more about betrayal. Moreover, when right-leaning people spoke about Australian society, they also spoke about a rejection of authority, whereas those who identified with far-right groups spoke more about respect for authority (e.g. reflected in topic 10, above).

Modelling when moral talk turns to talk of action in online interactions

We conducted a series of hurdle models using Mplus v8.10 to examine whether moral language was associated with collective action language across ideological group and issue. We used the score values from the eMFD which represent the prevalence of the language for each moral foundation (see Supplemental Table S5 for score values). The analyses distinguished between whether moral talk was associated with talk of action (i.e. as a binary variable; present or absent), and if so, by how much (i.e. as a continuous variable; degree of action language). Given the volume and complexity of the information generated by these analyses, we have provided summary plots of the effects for the different moral foundations for each of the binary and continuous predictors, for each of the ideological groups (see Figures 5 –7; see Supplemental Table S6 for specific regression values).

Figure 5.

Australian society and identity: moral language as a predictor of action language by ideological group.

Figure 6.

Climate change and the environment: moral language as a predictor of action language by ideological group.

Figure 7.

COVID-19: moral language as a predictor of action language by ideological group.

Australian society and identity

Figure 5 shows that, for both ideologically left- and right-leaning people, talk of authority/subversion was associated with a greater degree of talk of collective action within the context of Australian society. When engaging in online interactions about panel representation (topic 1), domestic violence (topic 8), and how crises are handled (topic 10) per the topic models, it seems that respect (or rejection) for authority and the government was central to the conversation for both ideologically left and right groups. This effect was strongest for those who identify with the far-right group, and notably, talk of authority/subversion was the only moral foundation associated with action for this group.

Nevertheless, there were notable differences between ideologically left and right people in their moral language as a predictor of action language. For instance, talk of acting was associated with talk of fairness/cheating for both ideologically left and right groups; however, the groups differed in the direction of this relationship: less action talk for right-leaning people but more action talk for left-leaning people. In addition, for left-leaning and ideologically neutral people (but not right-leaning people), talk of sanctity/degradation was uniquely associated with less talk of collective action. For the ideologically neutral group, sanctity/degradation was the only foundation that was associated with both the presence and degree of action language.

Climate change and the environment

Across all four ideologies, we found no instances where moral talk was reliably associated with both the presence and degree of collective action language when discussing climate change (see Figure 6). Nevertheless, among ideologically left, neutral and far-right groups, the presence of action talk was commonly associated with talk of authority. The topic model showed a focus on accountability (topic 2), evidence (topic 3) and political responsibility (topic 11) and suggests that, bound up with these discussions, was talk of authority and grassroots action.

Figure 6 shows that, while for left-leaning group members, talk of acting was predicted by a range of moral foundations (fairness, loyalty, authority and sanctity), for ideologically far-right people talk of protest was explained only by authority language. Talk of fairness/cheating was also associated with talk of acting for ideologically neutral groups. Strikingly, action talk among ideologically right people was not explained by moral talk.

COVID-19

As shown in Figure 7, talk of acting in the context of COVID-19 was explained by a common range of moral foundations (loyalty, authority and sanctity) for all four ideological groups. Specifically, for all ideological groups, talk of sanctity/degradation was linked with greater talk of action. In line with the topic models, action and shared concerns for loyalty, authority and sanctity were intertwined with COVID-19 conversations about political figures (topic 7), disease prevention (topic 12) and leadership (topic 13). These similarities in moral foundations endorsement were particularly prevalent for the right-leaning ideologies who both had the same pattern of association between moral and action talk.

However, more noticeable differences between left and right ideological groups arose regarding whether the care and fairness foundations predicted action. For ideologically neutral and right-leaning groups, but not left-leaning groups, their talk of fairness/cheating predicted a greater degree of action talk. The strength of this effect was strongest for those ideologically far-right. In contrast, care/harm talk was more often linked to action talk for ideologically left and neutral people, but not for right-leaning groups.

Discussion

The current paper sought to map a year of socio-political interaction in a modern town hall as an important setting for the regulation of social and political life within democracies. We examined the interplay between political ideology, moral foundations and collective action in the context of three different issues covered in a years’ worth of Australian political debate: Australian society and identity, climate change and the environment, and COVID-19. We found evidence supporting that town hall discussions about collective action to bring about a desired change were motivated by talk of moral values, but that both the quantity and quality of the moral language use differed across discrete ideological groupings (left, neutral, right, far-right) and socio-political issues.

Left and right: similarities and differences in how moral talk relates to action

Prior research using self-report methods would suggest that care- and fairness-based moral foundations would be most consistently associated with action for left-leaning people (Graham et al., 2009; Kivikangas et al., 2021). However, we found that variations of all five foundations are associated with talk of collective action when left-leaning people engage in socio-political interactions (see also Wang and Inbar, 2021), confirming prior work suggesting there is considerable heterogeneity in such ideological groupings in online interaction (e.g. Weber and Federico, 2013). Left-leaning people can be motivated to act from concerns for the welfare and rights of individuals (care and fairness foundations) and by the belief in groups and institutions which keep individuals in check (authority, loyalty, sanctity foundations; see also Milesi and Alberici, 2018). Thus, when left-leaning groups talk about moral values, they are also likely to talk about acting to protect or advance those positions (van Zomeren et al., 2023).

Traditional readings of MFT would suggest that more right-leaning or conservative people would be motivated to act collectively from a sense of loyalty, obedience to authority and purity (‘binding foundations’; Graham et al., 2009; Kivikangas et al., 2021). While ideologically right and far-right groups’ talk of acting was consistently explained by talk of authority, we did not find reliable support for the role of loyalty and sanctity, aligning with previous recognition that moral foundations play a role in some contexts more than others (Haidt and Joseph, 2011; Suhler and Churchland, 2011). Indeed, in the context of climate change, moral values were not found to motivate action for right-leaning people, even though such language was prevalent in their interactions (Supplemental Figure S4). That is not to say that morality is less relevant to right-leaning groups; rather, our findings suggest that in these data, online talk about morality did not coincide with talk of protecting the group’s moral values via collective action. Alternatively, it may be that engaging in collective action for climate change out of moral conviction is less relevant for right-leaning groups given they are less likely to believe in climate change (see meta-analysis by Hornsey et al., 2016). However, one limitation of measuring the presence, and not the nature, of action language means we cannot differentiate between whether groups were talking about acting to change or maintain the status quo (i.e. progressive vs conservative actions; Thomas and Osborne, 2022).

When far-right groups engage in town hall discussions about socio-political issues, it seems that talk of action is motivated by a narrower range of moral foundations than MFT would suggest (Graham et al., 2009; Haidt and Graham, 2007). For far-right groups, a moral value of authority/subversion was the only foundation associated with collective action when talking about Australian society and climate change. That is, far-right groups’ action was primarily motivated by concerns for leadership, government and tradition. This pattern is consistent with work by Alizadeh et al. (2019) who similarly found greater concerns for authority among far-right groups in online Twitter interactions, compared to ideologically right groups or left-wing extremists. We extend this work to show that a moral value of authority is important not only for discussions among far-right groups about socio-political issues, but it can also motivate talk of action in these contexts. This finding is perhaps unsurprising given that far-right parties are commonly in favour of systems which emphasise a respect for authority and discipline (Golder, 2016; Mudde, 2019). Notably, however, the sentiment of the language changed across contexts. For example, far-right groups were more likely to speak about a rejection of authority when discussing climate change. Far-right groups are often less likely to believe in climate change (Hornsey et al., 2016), suggesting they may have contested authorities who spoke about climate action and its role in the Australian wildfires. In contrast, a respect for traditional values and ways of life among far-right groups (Blee and Creasap, 2010) may explain a respect for authority when talking about Australian society and identity.

Nevertheless, concerns for authority were not unique to far-right groups. Rather, for all ideological groups, a moral value of authority/subversion was associated with collective action across all three contexts (with once exception of climate change for the right-leaning group). Throughout a year of political debate in Australia, talk of action was most often associated with talk of maintaining social order and respecting (or rejecting) government leadership. As such, people with different political orientations can be motivated to act based on the same moral foundation. Thomas et al. (2022b) similarly found that both left- and right-leaning groups in Australia used authority moral language when engaging in online interactions about a right-wing rally. Within an Australian context, talk of authority may be particularly relevant because democratic societies are typically concerned with the legitimacy of, and trust in, government institutions (Buchanan, 2002), especially in the online space where the spread of mis/disinformation can undermine democratic values and trust in the government (Dowling, 2022; McKay and Tenove, 2021). This goes to the heart of what collective action is within a democracy – it is to do with people seeking to change the societies of which they are part of through a dialogue with authorities. Indeed, across all three socio-political contexts there were recurring discussions about the democratic processes on the Q&A programme, including concerns for whether social groups were equally represented on the panel and general critiques of government leadership. These discussions touch on common concerns for procedural justice within democratic societies and highlight the role of town hall meetings in creating a platform for the regulation of socio-political life (Field, 2019).

Although moral talk of authority/subversion was common among the four ideological groups, the sentiment of this talk differed within the online interactions. In particular, the neutral group engaged in town hall discussions about Australian society, climate change and COVID-19 using more positively valanced moral language compared to the other groups. This finding may reflect that politically engaged groups (vs neutral groups) are more likely to be critical of and aggrieved over socio-political issues (e.g. by talking about cheating, betrayal, subversion). It is worth noting that while left- and right-leaning people differed – descriptively – relatively little, the associations between the moral talk and action talk were quite different. Given much of the literature on MFT and political ideology take a dichotomous approach to compare left- and right-leaning ideologies (Kivikangas et al., 2021; Suhler and Churchland, 2011), our findings offer new insights into centrist positions which are particularly relevant to an Australian setting where political ideology sits centre-left and centre-right (Bongiorno, 2022). Although we have adopted a macro, descriptive approach to focus on many elements simultaneously (moral foundations, political ideology, action, issue), future research could adopt a narrower focus to explore, for example, this difference in moral sentiment among centrist and partisan groups.

The role of context: moral values differ based on the issue

When considering the socio-political context of the online interactions, the interplay between political ideology, moral foundations and action was most distinct for discussions about COVID-19. While left-leaning people endorsed a wider range of moral foundations when discussing action for all three issues, for ideologically neutral, right and far-right groups this pattern was only observed when interacting about COVID-19. It seems that when engaging in town hall discussions about a deadly, existential threat, moral values mattered in general for collective action, regardless of ideology.

This finding aligns with other work on COVID-19 and moral foundations showing that morality is more salient when faced with disease threat (Henderson and Schnall, 2021) and vaccine controversies (Zhang et al., 2023). Indeed, recent theorising suggests that our understanding of relatively universal psychological phenomena may change as a function of COVID-19 (Henderson and Schnall, 2021; Rosenfeld et al., 2022). Overall, these findings demonstrate that talk of collective action within town hall settings is shaped by not only the type of moral talk, but also the socio-political context in which the moral attitude is held.

Limitations and future research directions

To map a year of political debate in a modern town hall setting, we focused on the Q&A programme as a distinct context where such debates take place. However, a limitation of the data is that the ABC viewership tends to be more left-leaning (Meade, 2023) and Twitter is/was also a more left-leaning platform (Pew Research Centre, 2020). This is reflected in our dataset where the ideology classifier identified our sample as primarily left-leaning. Of course, the ideology classifier itself is subject to a certain degree of error (Ram et al., 2025). An important direction for future research would be to collect data from a more right-leaning platform (or perhaps parliamentary records) and offer a comparison of the role of moral foundations, political ideology, action and context.

Second, LIWC has been extensively adopted in the social sciences (Boyd and Schwartz, 2021; Tausczik and Pennebaker, 2010) and the moral foundations and collective action plug-ins have been validated (Hopp et al., 2021; Smith et al., 2018). However, they have not been validated in an Australian context and the use of lexicon-based sentiment analysis in general is not without its limitations. LIWC can often over-identify expressions and, as a ‘bag of words’ dictionary approach, overlook context (Bantum and Owen, 2009; Lizzio-Wilson et al., 2025; Puschmann and Powell, 2018). Future research should account for these limitations by adopting a machine-based learning approach (e.g. Hoover et al., 2020).

Finally, the degree of talk about collective action throughout a year of political debate in Australia is probably underrepresented in these data. Given we collected these data in 2020 amid COVID-19, Australians were limited in their ability to engage in protests and other major offline gatherings due to social distancing rules and state lockdowns aimed at reducing the transmission of disease. The all-encompassing nature and threat of a global pandemic also means it is likely there were discussions about COVID-19 among the interactions about Australian society and climate change.

Conclusion

Throughout a year of political debate in an Australian modern town hall, talk of action was most often associated with talk of respecting (or rejecting) authority, across all socio-political contexts and all ideological groups. This signals the very heart of what collective action is: it represents a form of dyadic exchange between people who seek to bring about a particular social change, and the authorities who (in a democracy) are responsible for enacting it (Louis et al., 2020). Yet, differences also emerged within the discrete ideological groups. When left-leaning groups spoke about action, it was commonly motivated by a range of moral foundations, whereas talk of action by right-leaning ideologies was mostly motivated by a moral value of authority. Finally, associations between morality and collective action were more uniform across ideological groups in some socio-political contexts relative to others, such as COVID-19. Our findings confirm the critical role of the combination of moral positions, ideology and context in regulating socio-political life within communities.

Supplemental Material

sj-docx-1-nms-10.1177_14614448251413686 – Supplemental material for Mapping the moral domain in a modern town hall: Moral foundations’ effects on collective action conditioned by ideology and issue

Supplemental material, sj-docx-1-nms-10.1177_14614448251413686 for Mapping the moral domain in a modern town hall: Moral foundations’ effects on collective action conditioned by ideology and issue by Emily A Haines, Emma F Thomas, Christina Stothard, David Kernot, Morgana Lizzio-Wilson, Irene Caminschi, Rohit Ram and Marian-Andrei Rizoiu in New Media & Society

Footnotes

Acknowledgements

N/A

Funding

The authors disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This work was supported by the Defence Science and Technology Group, Australia.

Ethical considerations

This study was approved by the University of Technology Sydney Research Ethics Committee (approval no. ETH22-7064) on 8 May 2022.

Consent to participate

N/A

Consent for publication

N/A

Data availability statement

The data that support the findings of this study are available on request from the corresponding author, EAH. The data are not publicly available due to X’s terms of service, which do not allow for datasets to be made public, including via rehydration.

ORCID iDs

Emily A Haines

Emma F Thomas

Morgana Lizzio-Wilson

Marian-Andrei Rizoiu

Supplemental material

Supplemental material for this article is available online. All supplementary materials for this article – including code and validation analyses – are openly available on the Open Science Framework (OSF) at:

Author biographies

Emily A Haines is a PhD candidate in the College of Education, Psychology and Social Work at Flinders University. Emily’s research sits at the intersection of social and political psychology, and her PhD examines the conditions under which members of structurally disadvantaged and advantaged groups may work together to promote progressive social change.

Emma F Thomas, PhD, is a social psychologist in the College of Education, Psychology and Social Work, Flinders University, where she serves as Matthew Flinders Professor and Australian Research Council Future Fellow. Her research resides at the intersection of social and political psychology, concentrating on the psychological mechanisms that drive political engagement, collective action, and extremism.

Christina Stothard, PhD, is a computational behavioural scientist working as Senior Scientist Project Lead at the Defence Science and Technology Group. She is also Adjunct Senior Lecturer in the College of Education, Psychology and Social Work at Flinders University.

David Kernot, PhD, is a Senior Research Specialist at the Defence Science and Technology Group. David’s work broadly covers areas of identity intelligence, intelligence analytics, and information sciences.

Morgana Lizzio-Wilson, PhD, is a lecturer in Social, Political and Feminist Psychology at the University of Exeter and is Co-Director of Equality, Diversity and Inclusion in the Psychology Department. Morgana’s work is driven by the desire to understand how we can acheieve and maintain progressive social change, particularly in the context of gender equality.

Irene Caminschi, PhD, is a Scientific Advisor at the Defence Science and Technology Group. She is also an honorary Associate Professor of the Department of Microbiology and Immunology at the University of Melbourne.

Rohit Ram, PhD, works in applied science and research commercialization across computer science and cross-disciplinary domains. He holds a doctorate in online social behaviour and discourse modeling, examining how people express opinions and behaviors in digital social environments.

Marian-Andrei Rizoiu, PhD, is an Associate Professor leading the Behavioural Data Science lab at the University of Technology Sydney. His interdisciplinary research crosses computer and social sciences, blending psycholinguistics, digital communication and stochastic modelling to understand human attention dynamics in the online environment, the emergence of influence and opinion polarization.

References

ABC (2023) About the show: what is Q+A? Available at: https://www.abc.net.au/qanda/about (accessed 7 July).

Agostini

van Zomeren

(2021) Toward a comprehensive and potentially cross-cultural model of why people engage in collective action: a quantitative research synthesis of four motivations and structural constraints. Psychological Bulletin 147(7): 667–700.

Alberici

Milesi

(2018) Online discussion and the moral pathway to identity politicization and collective action. Europe’s Journal of Psychology 14(1): 143–158.

Alizadeh

Weber

Cioffi-Revilla

, et al. (2019) Psychology and morality of political extremists: evidence from Twitter language analysis of alt-right and Antifa. EPJ Data Science 8(1): 1–35.

Bantum

EOC

Owen

(2009) Evaluating the validity of computerized content analysis programs for identification of emotional expression in cancer narratives. Psychological Assessment 21(1): 79.

Bhattacharya

Srinivasan

Polgreen

(2014) Engagement with health agencies on Twitter. PLoS One 9(11): e112235.

Blee

Creasap

(2010) Conservative and right-wing movements. Annual Review of Sociology 36(1): 269–286.

Bongiorno

(2022) Did Australia just make a move to the left? The Conversation. Available at: https://theconversation.com/did-australia-just-make-a-move-to-the-left-183611

Boyd

Schwartz

(2021) Natural language analysis and the psychology of verbal behavior: the past, present, and future states of the field. Journal of Language and Social Psychology 40(1): 21–41.

10.

Boyd

Ashokkumar

Seraj

, et al. (2022) The Development and Psychometric Properties of LIWC-22. Austin, TX: University of Texas at Austin.

11.

Brookes

McEnery

(2019) The utility of topic modelling for discourse studies: a critical evaluation. Discourse Studies 21(1): 3–21.

12.

Buchanan

(2002) Political legitimacy and democracy. Ethics 112(4): 689–719.

13.

Cohrs

(2012) Ideological bases of violent conflict. In: Tropp

(ed.) The Oxford Handbook of Intergroup Conflict. New York: Oxford University Press, pp. 53–71.

14.

Dowling

M-E

(2022) Foreign interference and Australian electoral security in the digital era. Australian Journal of International Affairs 76(1): 40–56.

15.

Evans

(2014) A computational approach to qualitative analysis in large textual datasets. PLoS One 9(2): e87908.

16.

Feinberg

Willer

(2013) The moral roots of environmental attitudes. Psychological Science 24(1): 56–62.

17.

Field

(2019) Town Hall Meetings and the Death of Deliberation. Minneapolis, MN: University of Minnesota Press.

18.

Forsyth

(2013) Political Social Networking: The New Town Hall (ScholarWorks@GVSU). Allendale, MI: Grand Valley State University.

19.

Frimer

(2020) Do liberals and conservatives use different moral languages? Two replications and six extensions of Graham, Haidt, and Nosek’s (2009) moral text analysis. Journal of Research in Personality 84: 103906.

20.

Garten

Boghrati

Hoover

, et al. (2016) Morality between the lines: detecting moral sentiment in text. In: Proceedings of IJCAI 2016 workshop on computational modeling of attitudes. Available at: https://media.mola-lab.org/file/1702633610480-MoralityBetweentheLinesDetectingMoralSentimentinText.pdf

21.

Golder

(2016) Far right parties in Europe. Annual Review of Political Science 19(1): 477–497.

22.

Graham

Haidt

Nosek

(2009) Liberals and conservatives rely on different sets of moral foundations. Journal of Personality and Social Psychology 96(5): 1029–1046.

23.

Graham

Nosek

Haidt

, et al. (2011) Mapping the moral domain. Journal of Personality and Social Psychology 101(2): 366–385.

24.

Grover

Bayraktaroglu

Mark

, et al. (2019) Moral and affective differences in U.S. immigration policy debate on Twitter. Computer Supported Cooperative Work (CSCW) 28: 317–355.

25.

Haidt

Graham

(2007) When morality opposes justice: conservatives have moral intuitions that liberals may not recognize. Social Justice Research 20(1): 98–116.

26.

Haidt

Joseph

(2011) How moral foundations theory succeeded in building on sand: a response to Suhler and Churchland. Journal of Cognitive Neuroscience 23(9): 2117–2122.

27.

Henderson

Schnall

(2021) Disease and disapproval: COVID-19 concern is related to greater moral condemnation. Evolutionary Psychology 19(2): 14747049211021524.

28.

Hoover

Portillo-Wightman

Yeh

, et al. (2020) Moral foundations twitter corpus: a collection of 35k tweets annotated for moral sentiment. Social Psychological and Personality Science 11(8): 1057–1071.

29.

Hopp

Fisher

Cornell

, et al. (2021) The extended Moral Foundations Dictionary (eMFD): development and applications of a crowd-sourced approach to extracting moral intuitions from text. Behavior Research Methods 53: 232–246.

30.

Hornsey

Harris

Bain

, et al. (2016) Meta-analyses of the determinants and outcomes of belief in climate change. Nature Climate Change 6(6): 622–626.

31.

M-C

Pavlicova

Nunes

(2011) Zero-inflated and hurdle models of count data with extra zeros: examples from an HIV-risk reduction intervention trial. The American Journal of Drug and Alcohol Abuse 37(5): 367–375.

32.

Kivikangas

Fernández-Castilla

Järvelä

, et al. (2021) Moral foundations and political orientation: systematic review and meta-analysis. Psychological Bulletin 147(1): 55–94.

33.

Kosinski

Wang

Lakkaraju

, et al. (2016) Mining big data to extract patterns and predict real-life outcomes. Psychological Methods 21(4): 493.

34.

Lizzio-Wilson

Thomas

Wenzel

, et al. (2025) What could be? Depends on who you ask: using latent profile analysis and natural language processing to identify the different types and content of utopian visions. British Journal of Social Psychology 64(2): e12853.

35.

Louis

Thomas

McGarty

, et al. (2020) The volatility of collective action: theoretical analysis and empirical data. Political Psychology 41: 35–74.

36.

Lukensmeyer

Brigham

(2002) Taking democracy to scale: creating a town hall meeting for the twenty-first century. National Civic Review 91(4): 351–366.

37.

McKay

Tenove

(2021) Disinformation as a threat to deliberative democracy. Political Research Quarterly 74(3): 703–717.

38.

Meade

(2023) Leftwing audiences value ABC and SBS much more than rightwing viewers do, survey finds. The Guardian. Available at: https://www.theguardian.com/media/2023/jun/14/most-australians-recognise-abc-and-sbs-as-important-to-society

39.

Milesi

Alberici

(2018) Pluralistic morality and collective action: the role of moral foundations. Group Processes & Intergroup Relations 21(2): 235–256.

40.

Mimno

Wallach

Talley

, et al. (2011) Optimizing semantic coherence in topic models. In: Proceedings of the 2011 conference on empirical methods in natural language processing, Edinburgh, Scotland, UK, 27-31 July 2011. Association for Computational Linguistics, pp. 262–272.

41.

Mooijman

Hoover

Lin

, et al. (2018) Moralization in social networks and the emergence of violence during protests. Nature Human Behaviour 2(6): 389–396.

42.

Mudde

(2019) The Far Right Today. Hoboken, NJ: John Wiley & Sons.

43.

Mullahy

(1986) Specification and testing of some modified count data models. Journal of Econometrics 33(3): 341–365.

44.

Pauls

Shuman

van Zomeren

, et al. (2022) Does crossing a moral line justify collective means? Explaining how a perceived moral violation triggers normative and nonnormative forms of collective action. European Journal of Social Psychology 52(1): 105–123.

45.

Pew Research Centre (2020) Differences in how democrats and republicans behave on Twitter. Available at: https://www.pewresearch.org/politics/wp-content/uploads/sites/4/2020/10/PDL_10.15.20.twitter.dems_.reps_.fullreport.pdf

46.

Puschmann

Powell

(2018) Turning words into consumer preferences: how sentiment analysis is framed in research and the news media. Social Media + Society 4(3): 2056305118797724.

47.

Ram

Thomas

Kernot

, et al. (2025) Practical guidelines for ideology detection pipelines and psychosocial applications. Proceedings of the International AAAI Conference on Web and Social Media 19(1): 1630–1648.

48.

Reimer

Atari

Karimi-Malekabadi

, et al. (2022) Moral values predict county-level COVID-19 vaccination rates in the United States. American Psychologist 77(6): 743–759.

49.

Richardson

Conway

(2022) Standing up or giving up? Moral foundations mediate political differences in evaluations of BLACK LIVES MATTER and other protests. European Journal of Social Psychology 52(3): 553–569.

50.

Roberts

Stewart

Tingley

, et al. (2014) Structural topic models for open-ended survey responses. American Journal of Political Science 58(4): 1064–1082.

51.

Robinson

(2011) Town Meeting: Practicing Democracy in Rural New England. Amherst. MA: University of Massachusetts Press.

52.

Rosenfeld

Balcetis

Bastian

, et al. (2022) Psychological science in the wake of COVID-19: social, methodological, and metascientific considerations. Perspectives on Psychological Science 17(2): 311–333.

53.

Roy

Goldwasser

(2021) Analysis of nuanced stances and sentiment towards entities of US politicians through the lens of moral foundation theory. In: Proceedings of the ninth international workshop on natural language processing for social media, online workshop, 10 June 2021. Association for Computational Linguistics, pp. 1–13.

54.

Ryan

(2017) No compromise: political consequences of moralized attitudes. American Journal of Political Science 61(2): 409–423.

55.

Skitka

(2010) The psychology of moral conviction. Social and Personality Psychology Compass 4: 267–281.

56.

Skitka

Bauman

Sargis

(2005) Moral conviction: another contributor to attitude strength or something more? Journal of Personality and Social Psychology 88(6): 895–917.

57.

Skitka

Hanson

Morgan

, et al. (2021) The psychology of moral conviction. Annual Review of Psychology 72: 347–366.

58.

Smith

McGarty

Thomas

(2018) After Aylan Kurdi: how tweeting about death, threat, and harm predict increased expressions of solidarity with refugees over time. Psychological Science 29(4): 623–634.

59.

Suhler

Churchland

(2011) Can innate, modular ‘foundations’ explain morality? Challenges for Haidt’s moral foundations theory. Journal of Cognitive Neuroscience 23(9): 2103–2116.

60.

Tausczik

Pennebaker

(2010) The psychological meaning of words: LIWC and computerized text analysis methods. Journal of Language and Social Psychology 29(1): 24–54.

61.

Thomas

Osborne

(2022) Protesting for stability or change? Definitional and conceptual issues in the study of reactionary, conservative, and progressive collective actions. European Journal of Social Psychology 52(7): 985–993.

62.

Thomas

Duncan

McGarty

, et al. (2022a) Mobilise: a higher-order integration of collective action research to address global challenges. Political Psychology 43: 107–164.

63.

Thomas

Leggett

Kernot

, et al. (2022b) Reclaim the beach: how offline events shape online interactions and networks amongst those who support and oppose right-wing protest. Studies in Conflict & Terrorism. Epub ahead of print 13 March. DOI: 10.1080/1057610X.2022.2034222.

64.

van Zomeren

d’Amore

Pauls

, et al. (2023) The intergroup value protection model: a theoretically integrative and dynamic approach to intergroup conflict escalation in democratic societies. Personality and Social Psychology Review 28(2): 255–248.

65.

Wang

S-YN

Inbar

(2021) Moral-language use by US political elites. Psychological Science 32(1): 14–26.

66.

Weber

Federico

(2013) Moral foundations and heterogeneity in ideological preferences. Political Psychology 34(1): 107–126.

67.

Wright

Taylor

Moghaddam

(1990) Responding to membership in a disadvantaged group: from acceptance to collective protest. Journal of Personality and Social Psychology 58(6): 994.

68.

Zhang

(2018) Sentiment, richness, authority, and relevance model of information sharing during social crises – the case of #MH370 tweets. Computers in Human Behavior 89: 199–206.

69.

Zhang

Wang

Liu

(2023) Moral expressions, sources, and frames: examining COVID-19 vaccination posts by Facebook public pages. Computers in Human Behavior 138: 107479.