Information warfare: is the ‘battle of narratives’ enough to explain digital influence?

What is the ‘battle of narratives’ in the digital age?

In the digital age, the battle of narratives lies less in the attention it pays to narratives than in the blurring it allows between three distinct concepts: the symbolic organisation of plausibility, cognitive reception and behavioural causality. By conflating them, it risks transforming a strategic metaphor into a general theory of influence. By separating them, it regains its true purpose: to describe the conditions under which certain interpretations become available and resilient.

Narrative approaches are not inadequate because they are inherently weak. They become so when they are asked to produce something other than what they are capable of establishing. They reconstruct the symbolic and affective infrastructures of plausibility; computational methods reveal certain distributions within them; the study of reception and behaviour must still establish the effects. It is within this framework and not in the notion of an autonomous clash between narratives, that a rigorous analysis of digital influence can be constructed.

From the information war to ‘architectures of plausibility’

In the context of information warfare, monitoring systems generally remain organised around messages, accounts, keywords or identifiable campaigns. They accurately describe the content in circulation, but do far less to capture the structures that make this content recognisable and plausible to certain audiences.

In Understanding and Addressing Misinformation About Science, the US National Academies emphasise that the consequences of a piece of information depend not only on its dissemination, but also on the audiences it reaches and its interaction with pre-existing individual and structural factors: values, beliefs, identities, preferences, trust and social position. They further distinguish four levels of intervention: production, demand, distribution and appropriation. They conclude that none of these can be considered in isolation as the primary point of action.

This approach leads to a shift in the unit of analysis. Institutions should map not only the content, but also the oppositions, roles, implicit causalities and affects that persist despite the renewal of messages: hidden truth versus o s official discourse; dispossessed people versus elites; threat versus protection; revelation versus concealment.

Content may be removed or refuted whilst the architecture that lent it plausibility remains in place. Monitoring should therefore become longitudinal: it would track periods of dormancy, reactivations, changes in medium and contradictory appropriations of the same structure. The relevant object would no longer be merely the visible campaign, but the interpretative infrastructure capable of accommodating several successive campaigns.

The role of platforms: visibility, reactivation and distributed memory

Digital platforms are not merely channels that carry pre-existing content. On the contrary, they organise the visibility of narratives by selecting, prioritising and recommending certain formats over others. For example, algorithms, engagement metrics and short-form formats favour content that is recognisable, emotional and easily shareable. Thus, social media enables the reactivation of old content, which can resurface when an event restores its relevance. This logic creates a distributed memory, composed of fragments, reposts, montages, comments and memes. Narrative continuity therefore no longer depends on a single source, but on a multitude of actors and traces. Social media thus function as semiotic environments, where signs circulate, recombine and acquire new meanings. Understanding digital influence therefore requires analysing these technical, symbolic and affective environments, and not merely the messages they disseminate.

The role of emotions: why emotional coherence takes precedence over logical coherence

A narrative often seems plausible because it ‘resonates’ with a given audience. This resonance does not imply factual truth. It refers to the alignment between a proposition and a pre-existing framework of experiences, expectations and emotions. A local contradiction can then be absorbed without undermining the whole, provided that the emotional structure remains stable. Emotional coherence can compensate for weak logical coherence.

This property is reinforced by information overload¹. In environments characterised by speed, competition for attention and an abundance of signals, narrative structures serve to reduce complexity² . They assign roles, organise causal relationships and prioritise events. They enable interpretative continuity to be maintained without having to re-examine every piece of information. This cognitive economy explains their persistence, but it should not be immediately equated with persuasion.

Circulation, reception, behaviour: three things not to be confused

An analysis of narrative architectures naturally calls for a study of how they are received. An architecture may organise the perception of an event, provide a shared vocabulary or make an interpretation immediately available. Audiences may nevertheless share certain symbolic frameworks whilst retaining diverse judgements and behaviours. Discursive participation is generally less costly than practical mobilisation.

We must therefore distinguish between at least six stages: exposure to content, recognition of its structure, understanding, appropriation, endorsement and behaviour. The transition from one to the next is neither automatic nor linear. It depends on prior dispositions, group norms, the trust placed in the source, the costs of action, concrete opportunities and interactions with other information. The same framework may produce declared endorsement without behaviour, circumstantial behaviour without lasting endorsement, or mere familiarity.

Research into media effects³ has long demonstrated this dependence on the social contexts of reception. Narratives interact with broader environments; they do not constitute isolatable causes in the same way as a simple experimental treatment. Their role may be that of a framing device, a source of justification, a mechanism for reducing complexity, or a structure for emotional coordination. These functions are not equivalent and do not imply the same causality.

This distinction prevents us from concluding that a widely disseminated framework would necessarily be effective. Conversely, a significant behavioural shift may result from mechanisms that are not readily apparent: local normative pressure, material self-interest, organisational constraints or a triggering event. Narrative visibility and practical effectiveness may vary independently.

The limitations of the ‘battle of narratives’: from strategic image to causal illusion

The concept of the ‘battle of narratives’ possesses intuitive power and genuine strategic utility. It serves as a reminder that statements never circulate in isolation, that actors seek to impose frameworks of perception, and that control over meaning can be just as important as control over channels. It thus provides a common language for practitioners, analysts and researchers who previously observed related phenomena without always being able to link them together. This fruitfulness rightly explains its success.

Taken seriously, this metaphor even encourages us to move beyond the study of isolated messages. A battle cannot be reduced to the sum of the projectiles exchanged; similarly, a narrative space cannot be reduced to the sum of its contents. It possesses historical depth, positions already occupied, repertoires that can be mobilised, and audiences capable of recognising forms before examining their details. The concept therefore leads naturally to the analysis of architectures of plausibility discussed in this note.

Its limitation only becomes apparent when the strategic image becomes, almost without transition, a causal explanation. Speaking of a ‘battle’ suggests relatively coherent entities, identifiable positions and an attributable outcome. Digital environments are often more unruly: contradictory fragments may serve the same architecture, continuities may emerge without a single direction, and audiences may rework, reposition or treat the content presented to them with irony. This observation does not refute the metaphor; it simply indicates the point at which it needs to be supplemented.

What decision-makers should change

1. Stop describing mere circulation as ‘influence’

Institutions should stop using the volume of views, shares, comments or mentions as direct evidence of influence. These data measure exposure, circulation or reaction. On their own, they do not establish a change in beliefs, attitudes or behaviours.

This distinction ties in directly with the findings of the RAND study, Countering Violent Extremism in Indonesia, funded by the US State Department’s Global Engagement Centre. RAND notes that the evaluation of online campaigns still relies primarily on reach and engagement, likes, shares and comments, even though these indicators do not measure effects on attitudes or behaviour. The researchers therefore recommend the use of focus groups, before-and-after measurements and, where possible, comparisons between exposed and unexposed groups.

The Army Science Board adopts a similar approach when, in Human Interaction and Behavioural Enhancement, it recommends evaluating information operations on the basis of changes in online behaviour, the adversary’s reactions and observed effects on the ground, rather than solely on the basis of message dissemination.

A simple rule should therefore be applied: when a study has only platform data at its disposal, it should refer to visibility, circulation or responsiveness. The term ‘influence’ should be reserved for situations in which a cognitive, social or behavioural transformation has actually been documented.

2. Replacing message monitoring with the analysis of plausibility architectures

Monitoring systems are generally organised around messages, accounts, keywords or identifiable campaigns. They accurately describe the content circulating, but do far less to capture the structures that make this content recognisable and plausible to certain audiences.

In Understanding and Addressing Misinformation About Science, the US National Academies emphasise that the consequences of a piece of information depend not only on its dissemination, but also on the audiences it reaches and its interaction with pre-existing individual and structural factors: values, beliefs, identities, preferences, trust and social position. They further distinguish between four levels of intervention: production, demand, distribution and appropriation. They conclude that none of these can be considered in isolation as the primary point of action.

This approach leads to a shift in the unit of analysis. Institutions should map not only the content, but also the oppositions, roles, implicit causalities and affects that persist despite the renewal of messages: hidden truth versus the official line, the dispossessed people versus the elites, threat versus protection, revelation versus concealment.

A piece of content may be removed or refuted whilst the framework that lent it plausibility remains in place. Monitoring should therefore become longitudinal: it would track periods of dormancy, reactivations, changes in medium and contradictory appropriations of the same structure. The relevant object would no longer be merely the visible campaign, but the interpretative infrastructure capable of accommodating several successive campaigns.

3. Do not respond publicly to a narrative without assessing the risk of amplifying it

The institutional response must not be automatic. Publicly identifying a hostile narrative, exposing its origin or immediately issuing a denial may increase its visibility, confer upon it an importance it did not previously possess and reinforce the sense of persecution felt by the community disseminating it.

In Countering Russian Social Media Influence, RAND notes that publicly identifying a foreign source does not necessarily lead to a reduction in consumption, dissemination or belief. In some cases, exposing an operation may even amplify its message and give its author the strategic advantage of appearing particularly powerful.

In Health Misinformation, the US Surgeon General also recommends that information professionals learn to correct misleading information without contributing to its amplification, to respond first to the public’s genuine questions, and to prioritise credible local sources.

Before reacting, decision-makers should therefore weigh up at least four options: a public response, a targeted response, indirect action on the information environment, and no response at all. The decision should depend on the actual size of the audience affected, the extent to which they have internalised the narrative, the possibility of unintended amplification, and the existence of a documented strategic effect. An event that is highly visible but confined to a marginalised community does not necessarily justify a national intervention.

4. Stop funding counter-narratives without addressing what makes the opposing narrative plausible

Producing a ‘better narrative’ is not a sufficient policy. When a hostile representation is based on lived experience, an institutional contradiction or accumulated mistrust, the formal quality of the counter-narrative will not rectify the conditions that lend it credibility.

American research on the information ecosystem specifically shifts the focus towards relationships of trust, communities and institutions. In *Rebalancing the Information Ecosystem*, RAND emphasises that resilience is not merely a matter of disseminating accurate information: it involves restoring the social and institutional ties that enable that information to be received.

The US Army’s doctrine⁴ articulates the same requirement in more operational terms: a narrative must be reinforced by consistent actions, and the organisation’s behaviour must remain aligned with the message it conveys. Any narrative strategy should therefore begin with a plausibility assessment: what experiences, practices or contradictions lend credibility to the opposing narrative? The response might then include communication, but also a change in practices, verifiable evidence and the use of intermediaries who are genuinely credible to the target audience. An institution cannot, in the long term, use its narrative to counteract what its day-to-day behaviour helps to confirm.

5. Subject every narrative campaign to a counter-narrative test

No campaign should be rolled out across the board simply because its designers consider it persuasive or because it achieves good engagement results. It should be tested on several audience groups and compared with a baseline scenario.

The experiment should measure separately message recognition, comprehension, recall, internalisation, changes in attitudes and observable behaviours. It should also investigate any undesirable effects: amplification of the opposing narrative, increased polarisation, the reactivation of a long-standing issue, loss of trust or the mobilisation of an audience that was initially indifferent.

RAND specifically employed a controlled protocol to distinguish exposure to a campaign from its actual effects. Its researchers divided participants into a group exposed to the content under study and a control group, then measured the resulting differences. Whilst this method does not resolve all issues of ecological validity, it is far more robust than an extrapolation based on platform performance. The US Department of Defence places operations in the information environment within a broader framework combining personnel, programmes, governance and partnerships, confirming that they cannot be reduced to the mere production of messages⁵ .

Finally, institutions should require a critical analysis prior to deployment. A separate team would seek out counter-examples, unreceptive audiences, ironic interpretations and alternative causalities. This procedure would prevent analysts from selecting evidence that confirms the framework they themselves have constructed.

The analysis is available below

---

1. Jonathan Crary, 24/7: Late Capitalism and the Ends of Sleep, London, Verso, 2013. ↩︎
2. Yves Citton, Towards an Ecology of Attention, Paris, Éditions du Seuil, 2014; Bernard Stiegler, *The Automatic Society*, vol. 1, Paris, Fayard, 2014. ↩︎
3. Teun A. van Dijk, *News as Discourse*, Hillsdale, Lawrence Erlbaum Associates, 1988. ↩︎
4. Aguilastratt, A., Updike, M., & White, M. (2022). The Information Domain and Social Media: Part 2. NCO Journal. https://www.armyupress.army.mil/Journals/NCO-Journal/Archives/2022/January/Social-Media-Part-2/ ↩︎
5. Department of Defence. (2023). DOD Announces Release of 2023 Strategy for Operations in the Information Environment. https://www.war.gov/News/Releases/Release/Article/3592788/dod-announces-release-of-2023-strategy-for-operations-in-the-information-enviro/ ↩︎