The communication model of Shannon and Weaver in the digital age

Claude Shannon developed a mathematical theory of technical information transmission in 1948, and Warren Weaver expanded the perspective a year later to include questions about meaning and effect. For the analysis of digital communication, this distinction in the communication model of Shannon and Weaver remains relevant, because platforms influence which messages become visible, and artificial intelligence can now also be involved in their formulation.

From Prof. Dr. Patrick Peters, Professor of Communication and sustainability and Vice-Rector for Research and Teaching Material Development at the Allensbach University

The so-called Shannon-Weaver model is one of the most well-known representations of the communication process and often appears as a linear sequence: a sender transmits a message over a channel to a receiver; interference can occur along the transmission path. This representation is clear and simple, but it hides part of the model’s origin story. Claude E. Shannon did not develop a comprehensive theory of human communication in 1948. As a mathematician and engineer at the Bell Telephone Laboratories, he was concerned with the question of under what conditions signals could be transmitted as reliably as possible. Warren Weaver adopted this approach in 1949 and explicitly expanded it to include issues of the meaning and effect of communication. It was only the subsequent reception that consolidated both perspectives into a common communication model.

The separation of the two approaches is important, however, for the analysis of digital communication. At the technical level, signals are transmitted; at the semantic level, recipients interpret the transmitted symbols; only later can it be investigated what effect a message has. Shannon’s question therefore remains relevant for the technical infrastructure of digital communication. At the same time, digital media change the conditions under which messages are created, disseminated, and perceived. Platforms decide on the rules under which content is available, algorithms influence its selection and visibility, and generative AI can already be involved in its formulation. A linear relationship between a clearly defined sender and an equally clearly defined receiver only partially captures these processes. Shannon's model continues to describe the technical transmission, while additional theoretical approaches are required for the analysis of selection, interpretation, and effect.

Technical problem for Shannon's starting point

Claude Shannon published „A Mathematical Theory of Communication“ in two parts in 1948. Bell System Technical Journal. The essay is one of the foundations of modern information theory. Shannon formulated his problem with remarkable clarity: Communication at the technical level consists in reproducing an information selected at one point exactly or approximately at another point. He explicitly explained the meaning of the information as irrelevant to the technical problem (Shannon 1948). This statement is however sometimes misunderstood. Shannon did not deny the importance of semantic processes for human communication, but he methodically excluded them from the subject matter of his mathematical theory. He was interested in how a communication system must deal with possible messages whose selection before transmission is still unknown. Information in this theory therefore does not refer to semantic content in the everyday sense: the informational value depends on the selection of possible alternatives and their probabilities.

Shannon describes communication as a technical process of transmission. A source of information selects a message, a transmitter converts it into a signal, and this signal is transmitted over a channel. At the other end, a receiver reconstructs the message into its destination. During transmission, a noise source, referred to by Shannon as a „noise source“, can alter the signal. Shannon explains this model, among other things, using telephony and telegraphy. In telephony, the spoken message is technically translated into a transmittable signal and reconstructed at the other end of the connection.

With „noise,“ Shannon refers to disturbances that can interfere with the signal during this transmission. Technical noise can therefore lead to the receiver receiving a distorted signal. The mathematical communication theory examines, among other things, how reliably information can be transmitted over a channel and what consequences disturbances have for this process. Uncertainty and redundancy also play a role here, because they influence how much information can be transmitted under given conditions. However, technically flawless transmission still does not tell us whether the sender and receiver attribute the same meaning to a message. Likewise, it cannot be inferred from this what effect the message has on the receiver. Weaver made exactly these two questions the subject of his extension.

Weaver expands the technical problem to include its significance and impact

In 1949, Shannon's paper, along with Warren Weaver's „Recent Contributions to the Mathematical Theory of Communication“, was published as a book under the title The Mathematical Theory of Communication. Weaver placed the technical approach within a much broader context of communication. His definition of communication included methods by which consciousness influences another; in addition to language, he also included music, fine art, or theater. He even considered a further extension to the communication between technical mechanisms (Weaver 1949). Particularly fruitful was his distinction between three levels of problems. At level A is the technical question of how exactly communication signals are transmitted. Level B concerns the semantic question of how exactly the transmitted signals convey the intended meaning to the receiver. Level C concerns the effect: To what extent does the received meaning influence behavior in the desired way?

This distinction goes beyond the framework of sender, channel, and receiver. A message can be technically completely transmitted, but at the semantic level it still fails. Weaver himself points out that mere confirmation of understanding does not provide reliable evidence that both parties have actually understood the same thing. The semantic level thus concerns the difference between the intended meaning of the sender and the interpretation of the receiver. The third level introduces another aspect. For Weaver, communication is not limited to the agreement of meanings. Communication can have consequences for attitudes or behavior. At the same time, Weaver recognizes that semantic and effect-related questions are intertwined and analytically difficult to completely separate.

Therefore, it seems scientifically sensible not to prematurely merge Shannon and Weaver into a single model author. Shannon deliberately limited his theory to a technical transmission problem. Weaver saw in this theory a more extensive foundation and attempted to establish connections to semantic and effect-related questions. David Ritchie already pointed out in 1986 that both contributions should be considered analytically separately. He criticized in particular attempts to transfer Shannon’s mathematical theory directly to social science communication problems without examining its assumptions (Ritchie 1986).

Part of the model's success explains its problems

The so-called transmission model has been extremely influential in the fields of communication theory, management, and organizational research. Its appeal stems essentially from the fact that it breaks down communication into recognizable elements and makes visible where transmission problems can arise. However, difficulties arise when a model of technical information transmission is transformed into a complete description of interpersonal communication. Already in 1987, Bowman and Targowski criticized that the linear structure of complex human communication could only insufficiently capture it. The technical model assesses communication primarily from the perspective of transmission efficiency. On the other hand, interpersonal understanding arises in social situations where participants negotiate meanings and continuously react to one another (Bowman & Targowski 1987).

This criticism is also relevant to current teaching. In 2023, Getchell, Dubinsky, and Lentz examined the use of transmission models in textbooks on management and organizational behavior. Their study highlights the remarkable persistence of the Shannon-Weaver model in business education. There, the model continues to be a prevalent representation of communication, although its suitability for explaining complex organizational communication processes has long been debated (Getchell, Dubinsky, & Lentz 2023).

The boundary already emerges from Shannon’s initial problem. The message is present in his technical model before the transmitter encodes it. Human communication, however, can only produce meaning through exchange because participants react, ask questions, and adapt their statements to previous ones. Therefore, for digital communication, it would be less convincing to simply rename the classic components. A smartphone cannot be equated with the „channel“ in every respect, a social platform is not just an additional source of interference, and an algorithm is not a modern variation of technical noise. Behind these terms lie different processes that also need to be analytically distinguished.

Communication model according to Shannon and Weaver: Digital communication changes the function of the channel

For the transmission of digital data, Shannon’s technical theory remains directly relevant. However, in human communication over digital media, another question arises. A medium does not simply transmit a message under more or less undisturbed conditions. It also influences the possibilities that the sender and receiver have in formulating and interpreting a message. This can be particularly well observed in text-based communication. In an email or a messenger, many nonverbal cues are missing that are perceived in a personal conversation alongside the spoken language. For this reason, senders can revise their formulations before sending a message. Recipients, in turn, have time to read a message multiple times or to formulate their response only later. Such differences also arise even when the technical transmission works perfectly without errors.

Joseph Walther investigated these peculiarities of computer-mediated communication as early as the 1990s. His Hyperpersonal Model describes how users can more purposefully control their self-presentation under the conditions of text-based communication and how recipients develop impressions about their counterparts based on the available cues. These impressions can even be more idealized than in direct face-to-face encounters (Walther 1996). Walther and Monica Whitty later highlighted that the possibilities for selective self-presentation and time-shifted expression remain relevant even for newer forms of digital communication (Walther & Whitty 2021).

For the discussion with Shannon and Weaver, one thing is especially important: the channel can influence communication without the need for a technical disruption. A message can be transmitted exactly and yet understood under different conditions, because the medium used provides certain clues and obscures others. This is why Shannon’s concept of „noise“ is not sufficient for analyzing human communication in digital media. Weaver’s extension to the semantic level is particularly relevant at this point. As soon as the question is asked about how a receiver understands a message that has been transmitted technically correctly, the analysis moves from the technical transmission to meaning. Digital media change this process of meaning formation because they structure the communication situation differently than an immediate conversation.

Platforms are changing the relationship between the source and the recipient

Social media changes the relationship between source and receiver once again. Platforms provide the technical infrastructure and at the same time decide which contributions become visible. Algorithms sort content, prioritize certain contributions, and distribute them to different user groups. This adds an additional selection process between the publication of a message and its reception. The „Dynamic Intermediary Model“ („D[X]IM“) by Ohme et al. (2026) attempts to theoretically capture this constellation. The model assumes that communication on digital platforms is often not described as an immediate relationship between source and receiver. An intermediary intervenes in the information flow and influences which content is transmitted or made visible. This role can be assumed by, for example, journalists, influencers, artificial agents, or other actors within a platform. The dyadic relationship between source and receiver thus becomes a triadic structure.

A platform algorithm can, for example, deliver a post unchanged technically and still significantly influence who sees it at all. It thus does not generate any transmission errors in the Shannon sense. Instead, it changes the distribution of the message. Platforms use, among other things, behavioral and usage data to select content and individually control its visibility. Microtargeting also plays a role here: Different users can receive different communication offers based on available data. For the application of the Shannon-Weaver model, this distinction is central. Technical „noise“ alters a signal during transmission. Algorithmic selection begins earlier: It influences whether and under what conditions a message reaches the intended recipient. Therefore, algorithmic mediation cannot be meaningfully understood as a modern form of noise. It concerns the selection and distribution of communication and requires a separate theoretical description for this.

Artificial intelligence changes the question of the communicative actor

The involvement of generative AI in a message does not fundamentally call into question Shannon’s technical model. For the transmission, it is irrelevant whether a human forms a text alone or uses an AI system. What is crucial is that there is a message that can be technically coded and transmitted over a channel. However, communication science raises another question. When an AI system suggests formulations, adds content, or adapts a message to specific addressees, it is no longer readily discernible what proportion of the human user and what proportion of the system is involved in the actual message. Zhu, Markowitz, and Van Der Heide (2026) describe such forms of AI-mediated communication, among other things, using the terms „Co-Creation“ and „Adaptability“. For the technical transmission, little changes; for the attribution of authorship, intention, and responsibility, the difference can be significant.

The shift becomes even clearer when people interact directly with an AI system. In a chatbot, a human user is not interacting with a human receiver. The technical scheme of source of information, transmitter, channel, and destination can still be applied to the transmission. However, it does not answer the communication science question of whether and in what sense the AI system itself can be understood as a communication partner. The current AI communication thus primarily shows the boundary between technical and social description. Shannon can explain how a message is transmitted. Those who are supposed to be considered the author, addressee, or acting communication partner, on the other hand, belong to a different level of analysis.

Communication model according to Shannon and Weaver: „Noise“ is not a general term for all interference in the digital space

The very presence of digital communication suggests that the term „noise“ introduced by Shannon should be used more clearly. In everyday language, almost anything that hinders communication can be referred to as a disturbance. However, for a scientific analysis, such an extension leads to conceptual problems. Technical noise alters a signal along the transmission path. A semantic misunderstanding arises, however, when the receiver interprets the technically correctly transmitted message differently from what the sender intended. Algorithmic selection concerns the distribution and visibility of communication, and AI-supported text production can already change the very nature of a message. All of these processes influence communication, but follow different mechanisms.

Weavers three levels gain new relevance in this context. The technical question concerns the correct transmission. The semantic question concerns the interpretation of the message. The effect-related question concerns what the received meaning triggers. Digital platforms and AI bring further mediating processes into play that lie between these levels or precede them. The classic scheme therefore hardly lends itself to a meaningful update by simply adding another box for each new technical actor. Such an expansion would externally juxtapose different functions. More meaningful is the question of what an element actually changes: the emergence of a message, its technical transmission, its visibility, its interpretation, or its effect.

The distinction between levels is crucial for communication ethics

This differentiation also has an ethical dimension. Responsibility can only be meaningfully attributed when it is clear at which point in a communication process a problem arose. A technically incorrectly transmitted signal raises other questions than a misunderstood formulation. The situation is different when a platform influences the visibility of certain content or an AI system is involved in its creation. For communication ethics, therefore, it is crucial which actors actually influenced a process and what responsibility can be attributed to them. In digitally mediated communication, this responsibility can be distributed over several points: the formulation of a message, its technical delivery, its algorithmic selection, or its subsequent use. Shannon does not provide an ethical theory for this. However, his distinction between the technical level of semantic and effect-related questions helps to separate different areas of responsibility analytically. Weaver extends this perspective by explicitly treating meaning and effect as independent problems of communication.

What remains of Shannon and Weaver

The communication model proposed by Shannon and Weaver is particularly useful for human communication when its scope is precisely defined. It explains how a technical transmission process can be structured, and Weaver also makes it clear that successful technical transmission does not yet guarantee semantic understanding or a specific communicative effect. A digital message can be transmitted unchanged technically and still have a different meaning than intended by the receiver. On a platform, an algorithm can also influence its visibility. If generative AI was involved in its formulation, the communicative process begins even before publication. These processes can only be examined precisely if transmission, interpretation, selection, and production are analyzed separately.

This means that linear model drawing loses its claim to represent the entire communication process. Its analytical value remains intact, however. It compels us to treat transmission and understanding as distinct problems and to investigate the communicative effect separately once again. Under digital conditions, this distinction work has become more demanding. Platforms change the distribution of messages, while generative AI can already influence their linguistic form. The mathematical theory of Shannon does not explain such processes, but it does remind us to precisely determine the individual functions of a communication system before different phenomena are grouped under a general concept of communication.

In this sense, almost eight decades later, the approach still has scientific value. Shannon’s information theory is one of the technical foundations of digital communication, while Weaver expanded the perspective to include meaning and impact early on. Today’s research on computer-mediated communication, platforms, and artificial intelligence is a development from Shannon and Weaver, showing how much the constellations have changed since then and why the conceptual separation of the different communication levels remains necessary.

Literatur

Bowman, J. P. & Targowski, A. S. (1987): Modeling the Communication Process: The Map is Not the Territory. The Journal of Business Communication, 24(4), p. 21–34, https://doi.org/10.1177/00219436870240040.

Getchell, K., Dubinsky, J. & Lentz, P. (2023): A Critique of Transmission Communication Models in Introductory Management and Organizational Behavior Textbooks. Journal of Management Education, 47(5), p. 477–504, https://doi.org/10.1177/105256292311821.

Ohme, J., Mayer, A.-T. & Charlton-Czaplicki, T. / Gaisbauer, F. / Wedel, L. / Fan, Y. / Neuberger, C. (2026): D[X]IM—the Dynamic Intermediary Model of communicative transaction on digital platforms in a networked public sphere. Communication Theory, 36(1), pp. 12–23, https://doi.org/10.1093/ct/qtaf025.

Ritchie, D. (1986): Shannon and Weaver: Unravelling the Paradox of Information. Communication Research, 13(2), p. 278–298, https://doi.org/10.1177/009365086013002007.

Shannon, C. E. (1948): A Mathematical Theory of Communication. The Bell System Technical Journal, 27(3), p. 379–423; 27(4), p. 623–656. https://doi.org/10.1002/j.1538-7305.1948.tb01338.x and https://doi.org/10.1002/j.1538-7305.1948.tb00917.x.

Shannon, C. E. & Weaver, W. (1949): The Mathematical Theory of Communication. Urbana: University of Illinois Press.

Walther, J. B. (1996): Computer-mediated communication: impersonal, interpersonal, and hyperpersonal interaction. Communication Research, 23(1), pp. 3–43, https://doi.org/10.1177/009365096023001001.

Walther, J. B. & Whitty, M. T. (2021): Language, Psychology, and New New Media: The Hyperpersonal Model of Mediated Communication at Twenty-Five Years. Journal of Language and Social Psychology, 40(1), https://doi.org/10.1177/0261927X20967703.

Zhu, R. (C.), Markowitz, D. M. & Van Der Heide, B. (2026): Presenting the augmented self: extending sender effects of the hyperpersonal model in artificial intelligence-mediated communication. Annals of the International Communication Association (Advance Article), https://doi.org/10.1093/anncom/wlag029.