Understanding public opinion requires looking beyond headline polling and online reactions. This chapter presents approaches that help institutions uncover the values, experiences, concerns and perceptions that shape how people interpret policy issues. It outlines methods for continuous public listening, audience segmentation, engagement and discourse analysis, and explains how these insights can inform communication and policymaking alike. By developing stronger listening capabilities, governments can better understand the realities people face, identify space for compromise and anticipate points of resistance before positions become entrenched.
Public Communication Handbook to Bridge Divides and Shape Better Policy
3. Learning what people think beneath the surface
Copy link to 3. Learning what people think beneath the surfaceAbstract
Key takeaways
Copy link to Key takeawaysLook beneath surface‑level opinion. Deliberative polling, social listening, focus groups, and values‑based segmentation can reveal the deeper drivers of public attitudes – identities, emotions, lived experience, and the frames of reference people use to make sense of issues.
Treat listening as a continuous capability, not a one‑off exercise. Combining multiple inputs (public discourse, qualitative research, deliberative methods, and frontline insights) helps institutions detect early signals, identify misunderstandings, and anticipate points of resistance.
Understand how frames of reference and narratives shape interpretation. Mapping dominant frames and values across groups helps explain why similar messages land differently and why some narratives divide while others unite.
Use insights to guide policy, communication, and delivery. A structured “architecture of listening” channels what institutions learn – what matters, why it matters, and to whom – into agenda‑setting, policy design, communication strategies, and continuous adaptation.
People hold more complex views than simple polling can show. The goal of this toolbox is to help institutions understand what people really think about a policy problem beneath the surface of adversarial discourse, and to respond accordingly.
Effective public communication begins with a clear understanding of how people perceive, interpret, and react to policy problems. For governments, this means recognising that citizens are not passive recipients of information but active interpreters whose views are shaped by diverse values, identities, and lived experiences (OECD, 2021[1]; Macnamara, 2023[2]).
Understanding citizens and the drivers of their views and preferences should be the starting point of every communication activity. Likewise, it forms the basis of an evidence-based approach to policymaking: “the people factor” that needs to complement other evidence sources for sound policies. This is even more important when policy problems are value-laden, politicised, or involve difficult trade-offs.
Yet in practice, governments often tend to rely on fragmented inputs, one‑off consultations, or surface‑level opinion indicators that capture reactions to pre‑defined questions rather than the underlying concerns, values, and frames shaping public attitudes over time (OECD, 2025[3]; Macnamara, 2023[2]). This limits the ability of institutions to anticipate resistance, identify sources of friction, understand shifts in opinion, or adjust approaches before obstacles materialise.
Experience across OECD countries shows that public listening and engagement needs to move beyond episodic exercises and be treated as a continuous organisational capability (OECD, 2025[3]; OECD, 2023[4]). Communicators routinely monitor and analyse public discourse across open channels (e.g. social media) where citizens express concerns, expectations, and frustrations in their own terms. Unlike formal participation processes, this input is often open‑ended and continuous, allowing institutions to track how perceptions evolve over time and to capture early signals.
However, reliable listening effectively requires combining inputs and methods. Social media analytics, media monitoring, and sentiment tracking are useful to track narratives, but come with limitations (Martino et al., 2025[5]). These include risks of bias, over‑representation of vocal minorities, and dependence on opaque algorithms. On their own, these sources are insufficient proxies for citizens’ voices. They need to be combined with qualitative research, deliberative methods, and targeted engagement to understand why people hold certain views, how they weigh trade‑offs, and how identity, experience, and context shape their responses.
Listening and participation are complementary activities, but citizen participation1 methods go further: they serve to refine understanding and even to crowdsource parts of the policymaking (OECD, 2025[6]). Deliberative processes and similar approaches are the most effective way to identify preferences and find the space for compromise. Participation can also enhance legitimacy and build trust, but it must be meaningful: participation that is poorly connected to decision making, or that lacks visible follow‑through, can erode trust rather than strengthen it. The OECD has developed comprehensive guidance and compiled evidence of participatory processes that can support efforts to give citizens a say on decisions (OECD, 2024[7]; OECD, 2020[8]).
To address this, institutions can invest in organisational listening as a structured practice (Smillie and Scharfbillig, 2024[9]; Macnamara, 2025[10]; Macnamara, 2023[2]; OECD, 2023[4]). Rather than treating listening as a single method or tool, this approach views it as a system of interconnected processes, roles, and feedback loops that enable institutions to:
collect insights from multiple sources, including public discourse, research, participation processes, and frontline interactions
analyse and triangulate these inputs to distinguish signal from noise
channel insights into policy, service design, and communication planning
close the loop by explaining how public input was considered and why certain decisions were taken.
An “architecture of listening” (Macnamara, 2023[2]) helps institutions make listening intentional and actionable. It clarifies who is responsible for listening, how insights are shared internally, and how they inform decisions across the policy cycle.
This toolbox illustrates how to:
Make sense of public opinion on contested issues
Understand frames of reference and their evolution
Figure 3.1. The architecture of listening in an organisation
Copy link to Figure 3.1. The architecture of listening in an organisation
Source: Adapted from Macnamara, J. (2022[11]), Organisational listening in public communication: Emerging theory and practice, University of Technology Sydney, https://opus.lib.uts.edu.au/bitstream/10453/158330/2/Organizational%20Listening%20-%20Emerging%20Theory%20%26%20Practice.pdf.
3.1. Making sense of public opinion on contested issues
Copy link to 3.1. Making sense of public opinion on contested issuesUnderstanding public opinion today requires methods that go beyond traditional polling or routine media monitoring. Conventional tools provide useful snapshots and early signals, but on their own they often miss nuance, context, and the reasons why people hold certain views. Approaches such as deliberative polling, focus groups, and values‑based audience segmentation help communicators move beyond surface reactions to understand how people interpret issues, what values or experiences shape their views, and where room for compromise may exist. Combining methods allows institutions to distinguish between what people say, what they mean, and what ultimately matters to them.
Method choice should be deliberate and proportionate to the policy question, decision timeline and available capacity. Segmentation can help make sense of diversity within the public by identifying groups defined by shared values, worldviews or orientations, rather than demographics alone. New AI-enabled methods, such as synthetic sampling,2 can support survey design, test assumptions and explore possible reactions to policy options or messages, but they cannot substitute for empirical research or direct engagement with people. Used responsibly and alongside qualitative, quantitative and participatory methods, these approaches can help institutions develop a more accurate and actionable understanding of societal concerns while reducing the risk of drawing conclusions from partial or biased signals. The following questions can help guide the choice of opinion research methods:
What are the blind spots of my existing listening and opinion data?
How much of my data is quantitative versus qualitative? What does it tell me about audiences’ values, motivations, behaviours?
Which groups and perspectives do I most need to understand to manage risks and trade‑offs?
Do I need methods that capture reactions, or methods that help people reflect, deliberate, and explain why they hold certain views?
How confident am I that what I am hearing reflects broader public concerns rather than the loudest voices?
At what intervals do I need to update my data and analysis (for instance, which moments in the policy process)?
Table 3.1. Key methods to gauge public opinion and sentiment
Copy link to Table 3.1. Key methods to gauge public opinion and sentiment|
Research methods |
|
|---|---|
|
Social listening |
Analyses public conversations on digital platforms, forums and comment spaces to identify emerging topics and concerns. It is useful for tracking how issues are discussed in real time, but can overrepresent vocal, polarised or digitally active groups, so it should be triangulated with other methods. |
|
Surveys or polls |
Collect structured responses from samples of the population to measure awareness, attitudes, preferences or self-reported behaviours. Representative surveys provide broad snapshots and comparisons across groups or over time, but offer limited insight into why people hold certain views or how they weigh trade-offs. |
|
Deliberative polling |
Measures how public opinion changes when a representative sample receives balanced information and discusses an issue. Participants are surveyed before and after deliberation, making the method useful for understanding how informed reflection can shift views and reveal more considered preferences. |
|
Focus groups |
Use facilitated discussion with small groups to explore how people understand an issue, what concerns or motivates them, and how they react to policies or messages. They provide depth and nuance, but are not statistically representative and can be affected by group dynamics. |
|
Longitudinal surveys |
Track the same individuals or groups over time to show how attitudes, perceptions or behaviours evolve. They are useful for detecting trends and early warning signals, but require sustained investment and usually need qualitative research to explain why attitudes change. |
|
Synthetic sampling |
Uses AI-generated profiles or groups, based on existing data, to explore possible reactions to policies, scenarios or messages. It can help test assumptions or refine research design quickly, but synthetic samples are not real citizens and must not replace empirical research or direct engagement. |
Source: OECD authors’ elaboration.
3.1.1. Audience segmentation according to values and identities
Identity- and values-based segmentation of audiences is a method that uses representative population surveys combined with statistical clustering techniques to identify groups of people who share similar values, identities, and worldviews, rather than demographic characteristics alone. This approach, used notably by the non-profit More in Common3 across several countries and policy topics, recognises that people with similar socio-economic profiles may interpret issues very differently depending on their sense of belonging, trust in institutions, and underlying beliefs.
Quantitative segmentation (the size of the different clusters in the population) is complemented by qualitative interviews and focus groups, which deepen understanding of motivations, emotional drivers, and communication preferences. Together, these methods produce a set of recognisable audience segments that reflect how people see themselves and society, providing a more actionable basis for engagement and communication than demographics alone (see example in Box 3.1). Such segmentation can then inform efforts to use and support shared frames of reference that can offer a way into the debate for with most clusters. This approach helps practitioners to move beyond “average public opinion” by recognising that different groups interpret the same issue through distinct values, identities, and lived experiences.
Box 3.1. Examples of public opinion research methods
Copy link to Box 3.1. Examples of public opinion research methodsDeliberative Polling in the EU and Hungary
Deliberative polling, developed by the Stanford University Deliberative Democracy Lab, is a method to track on how public preferences evolve when citizens receive balanced information and engage in structured discussion. A representative sample is surveyed before and after being briefed by experts, and participating in moderated deliberation. For example, in 2009 a sample of EU citizens met to deliberate on key issues ahead of the European Parliament elections. Post-deliberation polls showed increased support for pursuing ambitious climate action even at economic cost (from 49% to 61%), while concern about immigration rose (from 44% to 64%). A similar exercise in Hungary in 2008 on labour policy saw support for increasing public spending on education, health and pensions funded by higher taxes rise from 26% to 34%.
Longitudinal surveys on public trust in Canada
The “Trust, Information, and Digital Ecosystems Study” (TIDES) is a national public opinion research programme that tracks how public attitudes, levels of trust, media use, and vulnerability to misinformation and disinformation evolve over time. It relies on repeated, nationally representative surveys to identify shifts in sentiment, early-warning patterns, and changing trust dynamics. By observing trends across time rather than at a single moment, the TIDES surveys support the early detection of shifts in beliefs, trust and distrust dynamics, media exposure and information pathways, responses to mis- and disinformation, and broader behavioural trends.
Values-based audience segmentation and clusters in Europe
More in Common’s (2025[12]) Europe Talks Migration research shows how identity‑ and values‑based segmentation can be applied to contested policy issues such as immigration. Across several European countries, this work demonstrates that public views on migration are rarely binary. Instead, people cluster into segments shaped by different values, emotional drivers, and perceptions of fairness and control. This approach helps communicators understand why similar headline positions may stem from very different concerns, and how engagement and communication strategies can be tailored accordingly. The table below illustrates what audience clusters look like in Germany.
Table 3.2. Example of audience clusters in Germany
Copy link to Table 3.2. Example of audience clusters in Germany|
Cluster |
Brief description |
Share of population |
|---|---|---|
|
The Open |
Value self-expression, open-mindedness and critical thinking |
16% |
|
The Involved |
Civic-minded and active democrats, value togetherness, and are willing to defend progressive social achievements |
17% |
|
The Established |
Value reliability and social harmony and are most likely to feel satisfied with the status quo |
17% |
|
The Detached |
Value success and personal advancement, are less likely to think in abstract societal terms or to be interested in politics |
16% |
|
The Disillusioned |
Have lost a sense of community and long for recognition and social justice |
14% |
|
The Angry |
Value order and control in national life, are angry at the system, and have very low levels of trust |
19% |
Source: Adapted from More In Common (2019[13]), Fault Lines: Germany’s Invisible Divides, https://www.moreincommon.com/media/0wih2zdp/more-in-common_fault-lines_executive-summary.pdf
Using AI to support the analysis of public consultation inputs: Greece
Greece's OpenGov.gr platform provides a structured channel for citizens, civil society organisations and other stakeholders to contribute to draft legislation and policy initiatives before their adoption. To strengthen its ability to process and learn from public input at scale, Greece is upgrading the platform with artificial intelligence capabilities that will help analyse, categorise and code comments submitted during consultations. The objective is to support the processing of large volumes of feedback and facilitate its use within the legislative process, helping policymakers identify key themes and insights emerging from citizen contributions.
Source: Deliberative Democracy Lab (2026[14]), What is Deliberative Polling®?, https://deliberation.stanford.edu/what-deliberative-pollingr#TheProcesshttps://deliberation.stanford.edu/what-deliberative-pollingr#TheProcess; Privy Council Office, (2024[15]). Trust, Information, and Digital Ecosystems Study (TIDES). Wave 8 data; More in Common (2025[12]), Europe Talks Migration: A cross-country study on changing public attitudes, https://www.moreincommon.com/europe-talks-migration/; Written contribution form the Government of Greece, September 2026.
A checklist for making sense of public opinion
Copy link to A checklist for making sense of public opinion1. Who is being heard?
Listening is ongoing and multi-layered, not based only on one-off or episodic inputs.
Evidence captures views across audience groups, including those that are underrepresented, harder to reach or less visible online.
Methods are accessible to different groups, including through translation, transcription and multiple ways to participate where needed.
2. What kind of insight is being generated?
Analysis goes beyond surface reactions to identify the values, identities, emotions and motivations shaping views.
Listening remains open to concerns raised by the public, not only to questions defined by the institution.
Findings are interpreted in context, including relevant historical, social, cultural and economic factors.
3. How robust is the evidence?
Methods are proportionate to the policy question, decision timeline and available analytical capacity.
Qualitative, quantitative and deliberative methods are combined where possible, with clear recognition of the limits and biases of each.
AI-supported tools are used transparently, responsibly and under human oversight, and synthetic or modelled insights are not treated as substitutes for direct engagement with people.
4. How are insights used?
Findings feed into communication strategies, policymaking and service design, rather than remaining standalone analysis.
Decisions document how public input was considered, including why certain views were or were not taken forward.
Changes in opinion are tracked over time to detect emerging risks, shifts in sentiment and delivery challenges early.
3.2. Understanding framings, narratives and their evolution
Copy link to 3.2. Understanding framings, narratives and their evolutionIdentifying and tracking dominant framings in public discourse on a contested topic sheds light on not only what people think about the issue, but how they make sense of it and talk about it. The methods discussed in this section build on those outlined above, complementing the analysis of public opinion by focusing on the interpretive frames through which issues are understood and debated. This analysis feeds directly into efforts to support shared, constructive frames of reference discussed in the next toolbox.
Frames of reference and narratives offer structured ways of making sense of reality: they shape how people interpret events, decide what matters, and orient their judgements (Dennison, 2021[16]). By conveying underlying values, assumptions, and causal beliefs, these framings influence attitudes and behaviours, particularly in contested policy areas (Seiger, 2025[17]; Dennison, 2021[16]). Understanding why frames and narratives carry such interpretive power in communication therefore requires distinguishing them from related concepts, such as stories or messages. Related terms are defined below in Figure 3.2.
A story outlines what happened
A message conveys the intended takeaway
A narrative is based on a framing and weaves multiple stories and messages together, giving events coherence and guiding audiences toward particular meanings.
Multiple, often competing frames of reference and narratives circulate at the same time, each seeking attention and legitimacy (Metropolitan Group, 2025[18]). Public communicators therefore operate in a crowded narrative environment that they need to understand before they can communicate effectively. This involves systematically mapping existing framings, identifying where they diverge or clash, and analysing the values, beliefs, fears, and expectations that make them resonate with different audiences. Only by doing so can communicators develop communications that connect more meaningfully with the public, strengthen trust, and support debate.
Keep in mind:
Frames of reference used in the media, online, and offline shape how people interpret new information, form opinions, and judge policy choices.
Frames and narratives are dynamic: they evolve across platforms, communities, and over time.
Narratives can be fabricated, but manipulated information spread not only through “bad actors” but also through legitimate grievances, emotional drivers, and social dynamics.
Understanding who creates, amplifies, and sustains certain frames of reference and narratives is as important as analysing the content itself.
Early, continuous monitoring helps anticipate challenges and respond before adversarial narratives become entrenched or permeate into mainstream discourse.
Figure 3.2. Key terms and definitions for analysing and mapping public discourse
Copy link to Figure 3.2. Key terms and definitions for analysing and mapping public discourse
Source: Authors’ work based on (Baú, 2016[19]; Badullovich, 2022[20]; Dennison, 2021[16]; Hajer and Versteeg, 2005[21]).
3.2.1. Methods to analyse and track existing frames and narratives
Understanding how frames and narratives form, evolve, and gain traction requires a combination of analytical approaches. Depending on available resources, time, and analytical capacity, communicators can draw on different methods to map dominant ones, track their evolution and salience over time, and understand why particular interpretations resonate with specific audiences (Table 3.3). No single method provides a complete picture. Narrative analysis is most effective when approaches are combined, allowing institutions to balance scale with depth, and to triangulate insights from public discourse, media, and direct engagement with people.
The output of this process can be conceived as a “discourse map” that for a given policy topic identifies distinct narrative communities and analyses how each group interprets an issue through its values, emotional orientations, and symbolic language. It breaks down these communities along with elements such as underlying values, sources of misunderstanding, and vulnerabilities to misleading or manipulative narratives. Box 3.2 below provides some examples of how these methods have been applied in practice.
Table 3.3. Methods to analyse and track frames and narratives in public discourse
Copy link to Table 3.3. Methods to analyse and track frames and narratives in public discourse|
Method |
Description |
|---|---|
|
Social listening |
Tracks how narratives emerge, spread and evolve across digital platforms. It identifies dominant themes, influential actors, framing patterns and amplification dynamics, while recognising that online activity may overrepresent vocal or polarised groups. |
|
Media content analysis |
Examines news articles, broadcasts and commentary to understand how issues are framed in traditional media, how visible they become, and how media actors shape public debate. |
|
Computational clustering and mapping (AI‑analytics) |
Uses AI or analytical tools to organise large volumes of discussion data into thematic narrative clusters, showing dominant topics, relationships between narratives and the wider structure of online discourse. |
|
Focus groups/community dialogues |
Engage members of the public directly to understand how they interpret issues, which values shape their views, and why some narratives resonate or fail. They help ground the discourse analysis in lived experience. |
|
Narrative workshops |
Use facilitated discussions where participants reflect on their values before responding to short narrative texts. They reveal which phrases, frames or story elements resonate, alienate or trigger disengagement. |
|
Literature scan |
Reviews academic research, policy reports and civil society analysis to map existing knowledge, contextual factors and the broader discourse around an issue. |
|
Expert interviews |
Gather insights from practitioners, journalists, researchers and civil society leaders with direct knowledge of the issue, including locally salient frames and how they connect to lived realities. |
|
Surveys |
Measure which narratives the public actually believes or references, helping distinguish narratives promoted by influential actors from those that genuinely resonate across the public. |
Note: This is a non-exhaustive list of methods. The labels used reflect those found in sources, but similar methods can have different names in different sources.
Box 3.2. Examples of discourse analysis methods and outputs
Copy link to Box 3.2. Examples of discourse analysis methods and outputsRe-Imagine Europa’s narrative maps
The narrative map method developed by non-profit Re‑Imagine Europa through its NODES initiative serves to identify distinct narrative communities that form around a policy issue. Although these communities respond to the same events, they interpret them through different assumptions, emotional orientations, and frames of reference. Narrative maps also reveal areas of shared concern among groups that may appear opposed and highlight which communities are more vulnerable to misleading narratives. A narrative map typically examines:
The narrative structure, including whether the situation is perceived as improving or deteriorating;
The values and priorities shaping each group’s interpretation, including which actors they see as responsible or legitimate;
The codes, symbols, and language that structure meaning within each group, noting that the same term, such as “freedom,” may carry very different connotations across groups.
Metropolitan Group multi-method analysis of democracy narratives
The non-profit Metropolitan Group (MG) applied a multi‑method framework to analyse pro- and anti-democracy narratives across 12 countries. The analysis combined literature review, expert interviews, media and social media monitoring, opinion survey review and political discourse analysis to map how narratives form, spread and connect with both public sentiment and elite messaging.
The analysis synthesised cross-country evidence into meta-narratives, coded them against core public values, and validated them through local-language research, surveys and focus groups in selected countries. This helped identify recurring anti-democracy narratives, such as “security justifies all”, and explain why they resonate across contexts by linking perceived threats to values such as security, strength, tradition and prosperity.
Mapping narratives with topic modelling and LLMs
Researchers at the NATO Strategic Communications Centre of Excellence (StratCom COE), supported by multilingual Natural Language Processing (NLP) specialists, used advanced AI tools to map how narratives formed across digital content in Estonian, Latvian, and Lithuanian, languages where standard AI models often perform poorly.
Resarchers combined open-source topic modelling, large language models and manually annotated Baltic-language text to map recurring themes and detect structured storylines across digital content. The narrative detection models produced structured interpretations of storylines, mapping how different pieces of content linked people, events, and motives into coherent narratives.
The AI-generated clusters and narrative interpretations were then reviewed by StratCom experts to assess whether they made sense in the cultural and political context of the Baltic States. This combination of automated large-scale analysis and human judgement helped identify narrative patterns that would be difficult to capture manually, while also showing where multilingual AI tools still struggle with local language nuance and context.
A checklist for understanding frames of reference, narratives and their evolution
Copy link to A checklist for understanding frames of reference, narratives and their evolution1. What framings and narrative are used in public discourse, and how are they moving?
Discourse analysis covers dominant, emerging and fringe frames and narratives across relevant channels, including mainstream, alternative, niche and harder-to-reach spaces.
Monitoring identifies where framings first appear and how they spread, evolve, merge or fade across platforms and over time.
Analysis distinguishes between visibility, salience and influence, rather than treating volume alone as evidence of importance.
2. Who is shaping and using the framings?
Analysis maps the wider discourse ecosystem, including producers, amplifiers, intermediaries, audience communities and network structures.
It identifies co-ordinated inauthentic behaviour, networked amplification, and tactics used by malicious or manipulative actors.
It assesses how credibility, trust and social positioning affect which actors can legitimise or spread particular narratives.
3. Why do frames or narratives resonate?
Analysis examines emotional, cultural and identity-based drivers that give narratives traction. It identifies the values, fears, hopes and grievances that make certain frames compelling to some.
It looks beyond factual accuracy to understand the conditions that allow inaccurate or misleading information to spread.
4. How robust and feasible is the analysis?
Qualitative sense making methods, are combined with quantitative tools.
Indicators of influence reflect credibility within communities, reach into relevant audiences, or potential for mobilisation, not just the number of mentions.
Findings are cross-validated across datasets and methods where possible.
The approach is proportionate to the time, skills and analytical capacity available.
5. How will insights be used?
Insights feed into communication strategies, policy planning and risk anticipation, not only reactive responses.
Findings are shared internally with other teams that can act on them.
Discourse maps are reviewed regularly and strategies are adapted when dynamics change.
6. What risks need to be managed?
Adversarial framings are not repeated unnecessarily or in ways that increase their visibility.
Safeguards are applied when analysing sensitive communities or topics.
Analytical limits are made clear, especially where intent, co-ordination or impact cannot be confidently established.
References
[20] Badullovich, N. (2022), “From influencing to engagement: a framing model for climate communication in polarised settings”, Environmental Politics, Vol. 32/2, pp. 207-226, https://doi.org/10.1080/09644016.2022.2052648.
[27] Barbu, E. et al. (2024), Narrative Detection and Topic Modelling in the Baltics, NATO Strategic Communications Centre of Excellence, https://doi.org/10.13140/RG.2.2.23252.56966.
[19] Baú, V. (2016), “Citizen engagement in peacebuilding: A communication for development approach to rebuilding peace from the bottom-up”, Progress in Development Studies, Vol. 16/4, pp. 348-360, https://doi.org/10.1177/1464993416663052.
[14] Deliberative Democracy Lab (2026), What is Deliberative Polling®?, Stanford University, https://deliberation.stanford.edu/what-deliberative-pollingr#TheProcesshttps://deliberation.stanford.edu/what-deliberative-pollingr.
[16] Dennison, J. (2021), “Narratives: A review of concepts, determinants, effects, and uses in migration research”, Comparative Migration Studies, Vol. 9/1, https://doi.org/10.1186/s40878-021-00259-9.
[29] Elreedy, D., A. Atiya and F. Kamalov (2023), “A theoretical distribution analysis of synthetic minority oversampling technique (SMOTE) for imbalanced learning”, Machine Learning, Vol. 113/7, pp. 4903-4923, https://doi.org/10.1007/s10994-022-06296-4.
[21] Hajer, M. and W. Versteeg (2005), “A decade of discourse analysis of environmental politics: Achievements, challenges, perspectives”, Journal of Environmental Policy & Planning, Vol. 7/3, pp. 175–184, https://doi.org/10.1080/1523908050033964.
[23] Health Information and Quality Authority (2024), Technical Report: National Engagement on Digital Health and Social Care, https://www.hiqa.ie/sites/default/files/2024-09/Technical-Report_National-Engagement-Digital-Health-Social-Care.pdf.
[25] Kapantaidakis, I. et al. (2025), “An Innovative Approach to Topic Clustering for Social Media and Web Data Using AI”, Computers, Vol. 14/4, p. 142, https://doi.org/10.3390/computers14040142.
[10] Macnamara, J. (2025), Public Communication and Public Policy: Reviving Democracy by Recalibrating Public Communication, Wiley Blackwell.
[2] Macnamara, J. (2023), Organizational Listening II: Expanding the Concept, Theory, and Practice, Peter Lang, New York.
[11] Macnamara, J. (2022), “Organizational listening in public communication: Emerging theory and practice”, https://opus.lib.uts.edu.au/bitstream/10453/158330/2/Organizational%20Listening%20-%20Emerging%20Theory%20%26%20Practice.pdf.
[5] Martino, E. et al. (2025), Ideological Fragmentation of the Social Media Ecosystem: From echo chambers to echo platforms, https://doi.org/10.48550/arXiv.2411.16826.
[18] Metropolitan Group (2025), Mapping the Global Democracy Narrative Landscape: Research Findings and Applications, https://www.metgroup.com/wp-content/uploads/2025/06/JUNE-2025-Mapping-the-Global-Democracy-Narrative-Landscape-DECK-Metropolitan-Group.pdf.
[12] More in Common (2025), Europe Talks Migration: A cross-country study on changing public attitudes, https://www.moreincommon.com/europe-talks-migration/.
[13] More in Common (2019), Fault Lines: Germany’s Invisible Divides.
[22] Neuendorf, K. (2017), The Content Analysis Guidebook, SAGE Publications, https://doi.org/10.4135/9781071802878.
[6] OECD (2025), Exploring New Frontiers in Citizen Participation in the Policy Cycle, OECD Public Governance Reviews, OECD Publishing, Paris, https://doi.org/10.1787/77f5098c-en.
[3] OECD (2025), Public Communication Scan of France: Evaluating public communication performance to strengthen public trust, OECD Publishing, https://doi.org/10.1787/7450361a-en.
[7] OECD (2024), OECD Survey on Drivers of Trust in Public Institutions – 2024 Results: Building Trust in a Complex Policy Environment, OECD Publishing, Paris, https://doi.org/10.1787/9a20554b-en.
[4] OECD (2023), Public Communication Scan of the United Kingdom: Using Public Communication to Strengthen Democracy and Public Trust, OECD Public Governance Reviews, OECD Publishing, Paris, https://doi.org/10.1787/bc4a57b3-en.
[30] OECD (2022), The Protection and Promotion of Civic Space: Strengthening Alignment with International Standards and Guidance, OECD Publishing, Paris, https://doi.org/10.1787/d234e975-en.
[1] OECD (2021), OECD Report on Public Communication: The Global Context and the Way Forward, OECD Publishing, Paris, https://doi.org/10.1787/22f8031c-en.
[8] OECD (2020), Innovative Citizen Participation and New Democratic Institutions: Catching the Deliberative Wave, OECD Publishing, Paris, https://doi.org/10.1787/339306da-en.
[28] OECD (2017), “Recommendation of the Council on Open Government”, OECD Legal Instruments, OECD/LEGAL/0438, OECD, Paris, https://legalinstruments.oecd.org/en/instruments/OECD-LEGAL-0438.
[15] PCO (2024), Trust, Information, and Digital Ecosystems Study (TIDES). Wave 8 data, Privy Council Office.
[17] Seiger, F. (2025), Navigating migration narratives - Research insights and strategies for effective communication, https://doi.org/10.2760/7999210.
[24] Shaw, C. and A. Corner (2017), “Using Narrative Workshops to socialise the climate debate: Lessons from two case studies – centre-right audiences and the Scottish public”, Energy Research and Social Science, Vol. 31, pp. 273-283, https://doi.org/10.1016/j.erss.2017.06.029.
[9] Smillie, L. and M. Scharfbillig (2024), Trustworthy Public Communications, Publications Office of the European Union, Luxembourg, https://doi.org/10.2760/69560.
[26] von Holstein, E. et al. (2024), The Power of Narratives - A Strategic Approach to Combatting Disinformation in Europe - Key Findings from the First European Narrative Observatory, Re-Imagine Europa, https://nodes.eu/wp-content/uploads/2024/11/NODES_WhitePaper_The-Power-of-Narratives.pdf.
Notes
Copy link to Notes← 1. The OECD Recommendation on Open Government uses the term “stakeholders” to refer broadly to both citizens and other interested or affected parties, while recognising that these groups should not be treated identically (OECD, 2017[28]). A stakeholder is defined as “any interested and/or affected party, including institutions and organisations, whether governmental or non-governmental, from civil society, academia, the media, or the private sector” (OECD, 2022[30]). Citizens can be defined as “individuals, regardless of their age, gender, sexual orientation, religious, and political affiliations. The term is meant in the larger sense of ‘an inhabitant of a particular place’, which can be in reference to a village, town, city, region, state, or country depending on the context. It is not meant in the more restrictive sense of ‘a legally recognised national of a state’. In this larger sense, it is equivalent of people (OECD, 2022[30]).
← 2. Synthetic sampling refers to the creation of artificial (“synthetic”) samples that replicate the statistical patterns and relationships of real data, allowing analysis when real samples are limited, sensitive, or unavailable. Based on (Elreedy, Atiya and Kamalov, 2023[29]), “A theoretical distribution analysis of synthetic data generation techniques”, Machine Learning, Vol. 113/7, pp. 4903-4923.
← 3. More in Common is an international research organisation that examines the social, political and cultural drivers of division in democratic societies. Its work focuses on identifying the factors that contribute to polarisation and declining trust, and on developing evidence‑based strategies to strengthen social cohesion. The organisation’s mission is to demonstrate that people share more common ground than is often reflected in public debate and to support policymakers, civil society and other partners in building more resilient, cohesive and open democracies. Based on: https://www.moreincommon.com/.