Jiang Junming: Challenges and Countermeasures for the Dissemination of Mainstream Ideology Under the Influence of Recommendation Algorithms
The assessment of the situation in the ideological struggle within cyberspace and the strategic deployment of related work constitute essential components of Xi Jinping Thought on Culture. General Secretary Xi Jinping has pointed out: "The issue of security risks in cyber ideology deserves high priority. The internet has become the current frontline of the ideological struggle. To maintain dominance over cyber ideology is to safeguard national sovereignty and state power. Party committees and member cadres at all levels must take the maintenance of cyber ideology security as a vital mission of 'guarding the territory and fulfilling one's duty' [1]. We must give full play to our institutional advantages, persist in the simultaneous application of management, utilization, and prevention, and ensure all parties take action to resolutely win the battle in the ideological struggle in cyberspace." By leveraging the technical advantages of selective manipulation—such as personalization, integration, and catering to preferences—recommendation algorithms have become a new paradigm for ideological dissemination and the shaping of values. This paradigm has enhanced the precision of ideological dissemination, but it has also amplified the limitations of individual thought and increased the density and disorder of the spread of various social trends of thought [2]. While making the dissemination of the mainstream ideology more difficult, it dilutes its authority and increases the security risks of cyber ideology, thereby pushing the struggle in cyberspace into a new realm. In his report to the 20th CPC National Congress, General Secretary Xi Jinping listed "strengthening the development of an all-media communication system and shaping a new landscape of mainstream public opinion" as an important component of "building a socialist ideology with strong cohesive and leadership powers." In his important instructions on propaganda, ideological, and cultural work, he further proposed: "Focus on enhancing the communication, guidance, influence, and credibility of news and public opinion." How to respond to the new circumstances in ideological dissemination brought about by recommendation algorithms, give full play to their technical advantages, control their attendant risks, and better disseminate mainstream ideologies—such as the basic principles of Marxism, fine traditional Chinese culture, and Xi Jinping Thought on Socialism with Chinese Characteristics for a New Era—is a major task requiring urgent resolution in the era of intelligence to "win the battle in the ideological struggle in cyberspace."
I. The Ideological Character of Recommendation Algorithms
A recommendation algorithm is an emerging technology capable of extracting network information and performing the matching and pushing of information based on user preferences. As a technology, the recommendation algorithm possesses no inherent bias or subjectivity. However, the process by which it is created and applied carries value positions and orientations. Its operating rules, processes, and links are an intersection of technical and human logic, reflecting at every turn human preferences regarding the selection of ideological content and value-oriented goals. The roles of "gatekeeper" and "recommender"—supported by recommendation algorithm technology through information filtering, processing, pushing, and reinforcement—endow it with a distinct ideological directionality and communicability.
As a product of the development of science and technology, recommendation algorithms are themselves tinged with ideology. Since the Industrial Revolution, science and technology have gradually become the dominant factor in the productive forces, and scientific and technological strength has become an increasingly important benchmark for measuring the degree of human social development. Driven by capital, the development of science and technology not only constantly changes people's thinking and behavior but has also gradually evolved into a product used to prove the "rationality" of bourgeois rule. It assumes the function of leading or even commanding ideology and thus possesses an ideological attribute. This was also an important factor in the decline of the socialist movement in the West after World War II. Herbert Marcuse, a representative figure of the Frankfurt School and an American scholar, pointed out in his book One-Dimensional Man that "the historical achievement of science and technology has rendered possible the translation of values into technical tasks." In other words, science and technology are produced within a certain social context; the "neutrality" and "rationality" they present are merely bestowed by rulers to maintain their current dominance. Science and technology always primarily serve the ruling class; under capitalist conditions, this means serving capital. Therefore, the ideological character of science and technology differs from traditional ideology; it relies on relatively mild means of presentation, utilizing technical products and technical thinking to permeate daily life, influence people's thoughts and behaviors, and modify their value orientations. Although the German scholar Jurgen Habermas did not agree with Marcuse's view that "the character of technology is political," he believed that "when technology and science penetrate the various institutions of society and thus change the institutions themselves, the old legitimations lose their validity," thereby indirectly acknowledging the ideological functions inherent in technology. While the relationship between technology and ideology still warrants in-depth study, it is an indisputable fact that technology in modern society serves ideology. Recommendation algorithms represent a new form of human information dissemination and reception characterized by revolutionary changes. However, this revolution has failed to change the political logic formed since the modern era—wherein science and technology are subordinate to capital—and has even further strengthened the function of science and technology as carriers of ideology that penetrate and influence human thought and behavior.
With the widespread application of recommendation algorithm technology, this technology has become deeply embedded in the socialization process of the populace, and its powerful capacity to carry values and shape ideology has become increasingly prominent. From the perspective of its operating mechanism, recommendation algorithms extract key variables such as the user (netizen), content, and environment to accurately depict "user portraits," quickly matching user preference information and performing targeted, high-density pushes. At the technical level, recommendation algorithms are merely tools for identifying, analyzing, and categorizing user information to provide data for a platform's information dissemination; however, at the ideological level, "every technical architecture, every line of code, and every interface represents a choice, implies a judgment, and carries a value." The determination of objects, the selection of data, the control of variables, and the priority levels for access all exude the air of value selection and are accompanied by the promotion and championing of ideology, thereby creating a social ecosystem conducive to the realization of certain goals and specific interests. A recommendation algorithm is nothing more than a technical method to cater to certain value concepts and a tool to serve a specific ideology. While this service appears on the surface to be centered on the needs and interests of the user, it essentially generates a more ingenious new form of ideological control and a new type of power for social control. This is the underlying logic behind the extensive application of recommendation algorithm technology in commerce and society.
First, the algorithm constitutes ideological control. Just as institutions have preferences, an algorithm is not only an operation at the technical level but also a calculation and norm at the social level. Goal anchoring for data, algorithm design, identification and filtering, integrated arrangement, and presentation of recommendations—the operating rules and system architecture inherent in recommendation algorithms—are all the work of technical personnel. Throughout the entire process of "data profiling," whether through intentional embedding (primary) or unintentional disclosure (secondary), recommendation algorithm technology is inevitably born under the shadow of the values of its technical creators. Of course, the social ecosystem behind the values of technical personnel is more a concentrated reflection of the groups and forces to which they belong; the content received by users will inevitably be profoundly influenced by cultural concepts, national interests, commercial capital, and even political power. The widespread application of recommendation algorithms allows their inherent thoughts and values to be spread to every corner of the network society, forming a power structure that intervenes in society and utilizes a subtle yet profound force to domesticate the thoughts and behaviors of users, thereby achieving an effect that serves "one's own" purposes.
Second, recommendation constitutes ideological guidance. Although algorithms push content based on the individual interests of users and influence the continuation, evolution, or direction of those interests, users do not possess true, autonomous "freedom of interest." This algorithmic treatment of interests and needs is merely a form of "analysis of learning conditions" [3] used in the process of recommending information to society for ideological dissemination. Its purpose lies in the guidance and shaping of value concepts; the "analysis of learning conditions" is used to improve the effectiveness of the recommendation. Furthermore, interest itself is the content of algorithmic recommendation; or rather, the recommendation algorithm itself includes the process of cultivating interest and user selection (filtering).
Finally, traffic contains the flow of ideology. Under the influence of instrumental rationality and market competition, recommendation algorithms often uphold the capital principle that "data and traffic are king." They are easily influenced and manipulated by capital forces and, if not handled carefully, can become tools for harmful forces to peddle their values. At the same time, people are prone to positive bias toward technology; faced with the intelligent and efficient information-pushing methods of recommendation algorithms, even if people do not understand the operating principles, they still choose to blindly trust and rely on them. Recommendation algorithms think and choose on behalf of the brain, constantly eroding people's capacity for independent thought and discernment between right and wrong, causing them to eventually become appendages of the algorithm through a process of "voluntary" "feeding." Thus, recommendation algorithms are reshaping people's cognition and behavior in an obscured manner, far exceeding the scope of technical instrumentalism and exhibiting a distinct ideological character.
In cyberspace, "the behavior of billions of people is constantly recorded, stored, and analyzed; the biological human is assigned digital attributes; the relevance of state/government power structures to each individual is weakening; compulsiveness is gradually diminishing; and individuals are re-parsed and values re-constructed on internet platforms." Recommendation algorithms typically reflect the contradictory unity of decentralization and re-centralization inherent in network society. Generally, decentralized pluralism is seen as an important feature of network society that distinguishes it from real-world society, and it is a key reason why it is favored by many: "There, anyone, anywhere, can express their beliefs without fear of being forced into silence or conformity, no matter how singular those beliefs may be." Recommendation algorithms take this a step further by inverting the relationship between information and people. Netizens are no longer passive recipients of information but are instead the selectors, the served, and the creators of information; it seems as though the needs and hobbies of netizens determine the generation and dissemination of information. Consequently, people may believe that all authority has become redundant, that network society does not need authority, and that there is no place to accommodate it. However, "in a sense, the development of the internet is a history of the constant struggle between decentralization and re-centralization." In reality, new technology has merely created a new domain for which powers compete: on one hand, recommendation algorithms reject the authority of state and government power or attempt to build a new public sphere outside of it; on the other hand, various political forces and capital platforms use algorithmic recommendation technology to peddle their own value concepts and expand their influence in more obscure and inflammatory ways, thereby winning the approval of more netizens (society) and gaining a foothold in the reconstruction of the ideological ecosystem. In the era of recommendation algorithms, ideological competition among various forces has become more intense, widespread, hidden, and complex. In a certain sense, whoever grasps dominance over recommendation algorithms for a given time grasps the initiative in ideological dissemination. This is increasingly vital for maintaining national ideological security and even political security.
II. Recommendation Algorithms Can Enhance the Effectiveness of Mainstream Ideological Dissemination
Mainstream ideological dissemination in China possesses broad coverage and intensive radiation power. However, from the perspective of the technical level and the dissemination methods it determines, characteristics such as low differentiation, one-dimensionality, time lags, repetitiveness, and excessive solemnity have led to unsatisfactory dissemination effectiveness. This has been a long-standing, difficult problem in mainstream ideological work, becoming even more apparent in a network society that emphasizes personalization. The most important function of recommendation algorithms lies in their influence on and alteration of the mechanisms of information allocation and dissemination. They can effectively expand the field of information dissemination and are becoming increasingly "understanding" of user needs and thoughts. The information provided increasingly "fits" user demands and even functions to provide "emotional value" [4], finding the intersection between the "meaningful" value positions of the mainstream ideology and "interesting" emotional needs. This provides a new type of technical choice for solving the problem of low dissemination effectiveness.
- Recommendation algorithms empower the precision of mainstream ideological dissemination. From data mining to tag classification, from information generalization to precision pushing, and from preference analysis to rapid feedback, recommendation algorithms provide "privately customized" services based on user needs and interests. On one hand, they draw user portraits to understand the user; based on big data platforms and using data and models as media, recommendation algorithms capture real-time data, correlate social behaviors, and clarify user habits, preferences, and preferred forms of reception. On the other hand, they rely on user portraits to predict the user: after gaining insight into preferences, habits, and needs, they perform targeted pushes of customized information, achieving a precise match between content supply and user demand and improving the precision of mainstream ideological dissemination. Recommendation algorithms take on the task of selecting the content and methods of mainstream ideological dissemination, helping the mainstream ideology shape the spiritual world and value concepts of users through subtle, imperceptible influence.
The communication and guiding power of ideology depend, to a certain extent, on the ability to control and utilize information technology. In fact, attention to political information is an important consciousness continuously cultivated throughout the process of socialization; coupled with the universal relevance of politics (and policy) itself, it can be said that nearly every netizen is a potential user of the mainstream ideology. However, the mass-scale and non-targeted modes of mainstream ideological communication have obscured the political interests of many people. Identifying information preferences and precisely locating the users corresponding to that information are the most prominent characteristics of recommendation algorithms.
Recommendation algorithms can effectively break through the limitations of time and space, analyze the needs behind users' online information, form user models and recommendation object models, and efficiently and accurately push useful information—not only in terms of choosing appropriate content but also in selecting the mode of engagement—to users. This effectively changes the inefficient practices of "flood irrigation" [5] and "high repetition" in the process of mainstream ideological communication, and avoids the situation where users feel lost or overwhelmed in a pan-socialized, fragmented, and disordered ocean of information. In terms of content, recommendation algorithms can use user data and models to precisely push information that interests them and meets their needs, such as various types of major principles and policies, laws and regulations, news events, contemporary commentary, political knowledge, historical analysis, and literary and artistic works. In terms of form, recommendation algorithms can analyze which discourse paradigms and information reception modes a user prefers—such as text, video, images, or music—through their behavior and choices. By selecting content and forms that are "loved by the masses" [6] to push information consistent with mainstream ideological values, they achieve personalized pushing and precision delivery, enhancing user acceptance and identification with the mainstream ideology. Forming a rich and diverse mainstream ideological "information pool" has become a basic condition for this work. Thus, recommendation algorithms effectively promote the attainment of a communication effect where mainstream ideology is "universalized without being generalized, and personalized without deviating."
2. Recommendation Algorithms Empower the Broadening of the Audience for Mainstream Ideology
Allowing as many citizens (users) as possible to become consumers of information, producers of content (reproducers of values), and subsequently disseminators of content (re-disseminators of values) is the core of how recommendation algorithms help broaden the audience of mainstream ideology. This entails enriching the mainstream ideological information pool through the innovation of communication forms and channels, expanding the audience groups of mainstream ideology, and cultivating assistants for its dissemination among the general public.
Recommendation algorithms assist in the innovation of mainstream ideological communication methods, which is conducive to manifesting the subjectivity of the audience and promoting the expansion and efficacy of dissemination. Under traditional communication models (including original online news dissemination), mainstream ideology exhibited characteristics of top-down, vertical, unidirectional communication. The audience was in a state of passive reception or even indoctrination, which neither guaranteed the effective acceptance of information nor provided opportunities for the expression of audience opinions. The communication efficiency was unsatisfactory, and the achievement of identification—especially deep-seated recognition—was difficult to guarantee. Under the influence of recommendation algorithms, horizontal (primarily in a formal sense) interactive communication methods enable interaction between New Media platforms for government affairs and users. Personalized, targeted, subject-oriented, and timely information-integrated network dissemination meets users' value requirements and needs for political participation. This effectively dilutes, conceals, or shifts the vertical characteristics (primarily in a content sense) of top-down value guidance, greatly reducing the discomfort and fatigue of netizens, inducing the audience's desire for interest expression and information creation following the mainstream ideology, effectively enhancing the sense of political participation efficacy, and boosting user identification with the mainstream ideology. Consequently, the communication power of mainstream ideology in network society also improves.
At the same time, recommendation algorithms can help enrich the channels of mainstream ideological communication, continuously expanding the audience reach and forming a situation of non-linear diffusion. At the technical level, by using function analysis to compare the similarity of users in the user pool and matching users with similar needs and interests, the algorithms leverage the advantage that "algorithmic technology possesses both diversity and individuality, as well as group and public attributes." This helps users break through temporal and spatial constraints to exchange and discuss mainstream ideological content they are interested in, expressing viewpoints and insights. This allows everyone to become a dynamic promoter of mainstream ideology, promoting interactive dissemination and expanding its influence and reach in the online sphere. For example, by using data such as "click-through rates," "dwell time," and "reposts" as standards to judge information focus, and setting up real-time "trending lists" for mainstream ideological topics, the algorithms capture users' curiosity and desire for the novel, effectively enhancing the breadth and influence of communication. By building online community platforms and giving full play to the grouping effect of communities, the themes of these communities can be designed to cover hot information within mainstream ideology, attracting interested users to enter the corresponding online communities for in-depth exchange and discussion. While deepening the sense of identification with the mainstream ideology, the "discussion heat" of the online community can further increase the popularity of the information, effectively expanding the scope of mainstream ideological dissemination.
As political communication (political propaganda) dominated by vertical directions shifts toward horizontal political communication among diverse political subjects, a balance is struck between users' value demands and the goals of mainstream ideological dissemination. While the number of users shows an upward trend, the expression of user experiences and the interactions between users constitute "non-linear multiple dissemination," presenting a new landscape of mass perception, group participation, and collective identification.
3. Recommendation Algorithms Empower the Real-Time Feedback of Mainstream Ideology
Currently, it is of little significance to judge the communication effect of mainstream ideology through statistical data such as subscription numbers or viewership ratings. Feedback on user reading and perception should instead serve as the basis for judging communication effects, which requires powerful technical support. Recommendation algorithms can obtain real-time feedback based on the user's reception status, providing the technical possibility for mainstream ideology to make timely adjustments to communication content, form, and timing, which is conducive to improving communication efficiency.
By monitoring users' browsing information, click-through rates, "likes," and comment content on platforms in real-time, and promptly drawing dynamic maps of user behavior, the recommendation algorithm's feedback mechanism gains important data support. On one hand, recommendation algorithms can use these dynamic maps to understand user psychology and attitudes, promptly correcting the push mechanism to achieve the optimal value for mainstream ideological communication effects. In the "pseudo-environment" [7] of mainstream ideological communication created by recommendation algorithms, users continuously receive corrected and supplemented new content, adjusting their cognitive attitudes and value orientations through a process of "subtle influence" [8]. On the other hand, dynamic maps are buoys for ideological security; analyzing the development trends of map curves can effectively identify ideological risks. Therefore, the real-time feedback of recommendation algorithms also acts as a real-time monitor for ideological dissemination. Once a trend toward public opinion risk or an ideological crisis appears, timely adjustments and precautions can be made, thereby "nipping the risk in the bud" and avoiding the uncontrollable fermentation of public opinion. Under the real-time feedback mechanism of recommendation algorithms, the effectiveness of mainstream ideological communication will significantly improve, and the probability of ideological risks will be greatly reduced.
As an emerging internet technology, recommendation algorithms have injected "fresh blood" into mainstream ideological communication, reshaping its mechanisms and methods in terms of precision content pushing, deep audience expansion, and real-time information interaction. However, in the specific application of new technologies, "rejection reactions" and delays in supporting measures always occur. The ability of relevant Party and government departments to use the real-time feedback functions of recommendation algorithms is still insufficient, and the road to truly realizing the effective integration of recommendation algorithms and mainstream ideological communication remains long and difficult. More importantly, in today's increasingly complex ideological struggle, recommendation algorithms have brought about huge changes in the social ecological environment and power structures, greatly increasing the variables in online ideological dissemination. This further induces the breadth and intensity of the spread of various Western social trends, false information, and useless information, as well as the risk of group thought polarization. Recommendation algorithms also have the potential to become accelerators of thought polarization, causing the dissemination of mainstream ideology to face challenges.
III. Challenges of Mainstream Ideological Communication Under the Influence of Recommendation Algorithms
General Secretary Xi Jinping has emphasized that "the internet has become the main battlefield of the struggle for public opinion. Some comrades say that the internet is the 'greatest variable' we face; if not handled well, it could become a 'heartfelt concern' [9]." The impact of recommendation algorithms on mainstream ideology is complex, and this complexity is rooted in the dual attributes of recommendation algorithms: technology and power. That is to say, this complexity goes far beyond technology itself; we should pay more attention to and emphasize the changes in the social power structure brought about by technology. Furthermore, although recommendation algorithms provide a new tool for Party and state media, thereby providing more powerful technical support for improving the efficiency of mainstream ideological communication, their widespread application has also brought a strong sense of urgency, pressure, and crisis to mainstream ideological work. This is because "algorithm as power"—this power hatched with the help of the network society—possesses the characteristics of "diffusion" and "accumulation" coexisting and mutually transforming. As the application scenarios and "diffusion" range of "algorithmic power" continue to expand, the challenges to information dissemination and control methods from "technological empowerment" are also growing. Various social and commercial platforms supported by recommendation algorithms are continuously competing with mainstream media for (and cultivating) online space and users. By supplying a large amount of information content such as gossip, entertainment, consumption, daily life, novelty, and rebellion, they construct various new types of online communities and capture more of the public and their reading time. This creates conditions for the public to become alienated from "grand, formal, serious, lofty, and mainstream" ideologies, even forming "circle-layer" [10] structures, thereby—intentionally or unintentionally—continuously weakening the Party's leadership over ideology and hollowing out the influence and identification of mainstream ideology in society.
1. Algorithmic Gatekeeping Narrows the Discourse Power of Mainstream Ideology
The American sociologist Kurt Lewin first proposed the concept of the "gatekeeper" in his book Channels of Group Life, which was later introduced into the field of journalism and communication studies. A "gatekeeper" refers to an individual or institution in mass communication that can decide the nature of the information to be disseminated, how much is disseminated, and how it is disseminated. Information gatekeeping is extremely important in the process of social communication, exerting a profound influence on the direction and value norms of information dissemination. Therefore, whoever holds the power of gatekeeping information resources gains the discourse power of ideological dissemination and can reshape the ideology of society. Under the traditional information dissemination model, the role of "gatekeeper" was undertaken by senior journalism professionals such as editors and reporters (who often hold political identities as Party members or cadres and mostly carry out their work under the leadership of primary-level Party organizations). They shoulder the important responsibilities of promoting mainstream values, correctly guiding public opinion, filtering out dangerous information, and maintaining social stability, promoting the status of mainstream ideology as the basic principle and theoretical basis for people in perceiving, explaining, and analyzing various social phenomena. After recommendation algorithms entered the field of information dissemination, the authority of "information gatekeeping" was largely ceded to so-called "value-neutral" algorithms and machines. To a certain extent, this avoids subjective bias in auditing, but "neutrality" is actually a type of value orientation as well, and behind it lies the logic of capital, which follows the principle that "traffic is king." The objectivity brought by mechanization weakens the positive guiding function that "gatekeepers" should have. While providing users with an "accommodating" experience, it also leaves users confined within a "technological cage."
As a scientific technology widely applied in the commercial and social spheres, recommendation algorithms have led to "major changes in the roles and status of the state, society, media platforms, and audiences in the field of information dissemination," presenting the mainstream ideology with a challenge to its discourse power. In the process of gatekeeping, recommendation algorithms follow instrumental rationality; the soaring of quantitative indicators such as "traffic," "frequency," and "subscription numbers" is the goal they pursue, serving as the optimal solution to all problems. This has incubated information push modes that cater to users, such as "Guess What You Like," "Topics of Interest," and "My Hot Searches." To gain more attention, likes, and traffic, algorithms even use selectively filtered data to fabricate a certain kind of "truth." Content related to the mainstream ideology is frequently placed in the user's "not interested" column, becoming an object to be screened out. Under the "fueling of the flames" [11] by recommendation algorithms, users only see information they "want to see" and ignore content they "ought to see" or "should see." This causes users to be indulged in a single, one-sided information environment for a long time, making them prone to cognitive fixation and ideological polarization, which brings tremendous trouble to the dissemination of the mainstream ideology. Faced with this passive situation, and based on considerations of political performance and management efficiency, party and government departments often resort to "vertical" bureaucratic leadership structures to increase the intensity of mainstream ideological dissemination (which often manifests as repetition in content and monotony in form). However, the result is a further "solidification" of the political and coercive nature of mainstream ideology. Not only does the effectiveness of dissemination fail to improve alongside increased intensity, but it can also trigger a "reactionary" [12] mentality among some members of the public toward the mainstream ideology. Meanwhile, while the "private customization" service of recommendation algorithms brings convenience to users, it also directs the focus of information dissemination toward novel, grotesque, and highly controversial topics—even utilizing fake headlines as gimmicks, which violates public order and good customs and disrupts social order. Alternatively, these algorithms "keep a distance" from grand, formal, serious, sublime, and mainstream topics and news, losing the pursuit of mainstream ideology and the adherence to social responsibility. They lose the value-based commitment of the "gatekeeper," thereby dissolving the discourse power and acceptance of the mainstream ideology.
Naturally, the practical orientation of recommendation algorithms is not satisfied merely with catering to user preferences; the underlying commercial logic, profit motives, and political intentions are the keys to controlling content delivery. It is evident that completely surrendering the power of information gatekeeping to recommendation algorithms cannot fulfill the task of verifying facts and discerning information. This leads to a narrowing of the channels for disseminating ideal beliefs and the public spirit, as well as a lack of content, meaning the mainstream social ideology lacks the necessary conditions and important foundation required for its widespread dissemination.
2. The "Technical Black Box" Weakens the Dominant Power of Mainstream Ideology
The operating mechanisms of recommendation algorithms embedded in various commercial and social platforms are often opaque. The transition from data input to information output is not a linear process. Users remain unaware of what value judgments and value purposes the information pushed to them is based on, and it is difficult for the relevant ideological departments to exercise control. The entire operational process seems wrapped in an imperceptible "black box," constituting a "technical black box." The characteristics of hidden generation and open functioning in recommendation algorithms provide numerous opportunities for the infiltration of erroneous trends of thought and harmful ideologies, thereby greatly weakening the dominant power of the mainstream ideology in society.
The "technical black box" principle followed by recommendation algorithms is, in itself, a negation of the dominant status of the mainstream ideology. The difficulty in popularizing the technical knowledge of recommendation algorithms and the associated costs mean that most users' understanding remains on the surface. As man-made artifacts, algorithms—intentionally or unintentionally—incorporate the ideological biases of technical personnel or the capital behind them. This invisibility and manipulability provide a hiding place for dangerous trends of thought and erroneous ideas. Ambiguous stances lead to value biases, whereby the presence and visibility of the mainstream ideology are continuously weakened. Furthermore, the information asymmetry caused by the technical black box exacerbates algorithmic bias, erodes user trust in push mechanisms, and fosters a questioning attitude toward pushed content, creating a distrustful online environment. This questioning is often first directed at the mainstream ideology, thus affecting its effectiveness in providing value leadership.
The liberalized nature of cyberspace increases the risks brought by the technical black box, encroaching upon the dominant strongholds of the mainstream ideology and causing online public opinion to present a state of "confluence." Whether it is the algorithmic collection of user information or the recommendation and supply of information, it is a state of technical black box that is almost undefended. The liberalized cyberspace presents a form of virtuality and anonymity, further expanding the dissemination field for various non-mainstream ideologies. In virtual cyberspace, the threshold for expressing opinions is lowered, and the harmful impulse to irresponsibly express one's own views is further stimulated by recommendation algorithm technology. This is followed by endless verbal battles, and a massive emergence of uninhibited ideological dissemination based on venting emotions without rational thought or considerations of traffic. "People's thinking appears to be in a gentle 'cage,' invisibly subjected to a greater degree of influence and discipline from capitalist consciousness, causing the authority of the mainstream ideology to be eroded and dissolved, posing an endogenous challenge to the dominant power of the mainstream ideology." Meanwhile, mainstream media and regulatory departments are either indifferent to the various contents pushed online and the main ideas they convey, or they find their strength lacking, or they simply resort to total bans. This accelerates the deterioration of the dissemination ecosystem for the mainstream ideology.
Ideological security is the key to national political security and the guarantee of social stability. However, the technical black box provides an imperceptible "invisible space" for the penetration of various harmful trends of thought. Certain forces with ulterior motives even use the names of so-called "revealing secrets," "objectivity," and "restoring the truth" on platforms to promote "historical nihilism" [13]. Through recommendation algorithms, they amplify and spread these views, gradually entering online public opinion, affecting the formation of values among youth, dissolving people's sense of national pride, and fragmenting national cohesion. Against the backdrop of the accelerated evolution of the "changes unseen in a century" [14] and the increasingly fierce collision between different institutional forms, the open and hidden struggles between ideologies emerge incessantly. Once controlled by capital or Western trends of thought, recommendation algorithms could become powerful weapons for anti-China forces to conduct cultural infiltration and ideological capture of our country, dissolving the people's identification with and confidence in the national path, theory, system, and culture, and threatening our country's ideological security.
3. Audience-Centricity Dissolves the Level of Attention Toward Mainstream Ideology
While customizing personalized information for users, the intelligent push mechanism of recommendation algorithms continuously promotes the prevalence of consumerism, utilitarianism, and "fragmentation-ism" [15] (distancing oneself from the sublime, the grand, and the intellectual). Non-mainstream ideologies and online subcultural content are showing a trend of becoming "mainstream" in cyberspace. Attention resources online are scarce resources that all ideologies must fight for; their limited and exclusive nature determines their importance in the field of communication. The higher the level of ideological attention, the more user attention resources it has obtained. The practice of recommendation algorithms selecting push content and methods based on user preferences and habits contains an audience-centered value of algorithmic communication. In contrast, mainstream ideological communication needs to be centered on the communicator and the communication content. Thus, it falls into a dilemma where content supply and user demand cannot fully coincide. In the context of information overload, mainstream ideological communication often cannot withstand the impact of interesting or entertaining content. It is difficult to obtain attention resources by relying on the spontaneous choices of netizens, as users tend to place their attention on information blocks they find more interesting.
Accompanying the segmented attention resources is the marginalization of the mainstream ideology. Under the "care" of the intelligent push mechanism, the user's standard for evaluating information becomes "is it useful?" Fragmented information scatters their time and energy; in the absence of steps for reflection and exploration, the user's perception of mainstream ideological content also shifts from "understanding" [16] to merely "being aware of." Serious and systematic mainstream ideological discourse is ignored, while a massive amount of pan-entertainment and fragmented information floods the user's brain, resulting in the construction and dissemination of mainstream ideology becoming fragmented, solidified, and disordered. The dominance of online subculture and non-mainstream ideological discourse creates growth space for erroneous trends of thought and improper concepts, secretly obstructing or even deconstructing the process by which mainstream ideology is internalized in users' hearts and externalized in their behavior.
The excessive entertainment of online information is often the product of collusion between commerce and capital and, in turn, boosts this collusion. Therefore, hiding behind vulgar and lowbrow content is the hand of capital. Under the "anesthesia" of entertainment content, users easily become addicted to the revelry of "low-level tastes" and become self-important. The dissemination and guidance of mainstream ideology thus appear even more difficult. Under the restrictions of "platform-first" rules, if mainstream ideological content is to use the platform as a springboard for dissemination, it must comply with the platform's rules and logic. This, in turn, exposes the mainstream ideology to the risk of decentralized deconstruction or results in the infiltration of alienated content, reducing the attention paid to the mainstream ideology and weakening the users' sense of identification with its authority and discourse power.
4. "Information Cocoons" Reduce the Integrative Power of Mainstream Ideology
The "Information Cocoon" is a problem that emerges alongside recommendation algorithms. People immersed in an information cocoon only listen to things they choose and things that please them. "Due to the overlapping effects of social networks and algorithms, the probability of people encountering information contrary to their own views on social media is only 5%–8%, and only a small portion of users will click on such content." The personalized information delivery and catering to preferences of recommendation algorithms construct a narrowing and fragmented state of information reception for users. Instead of enjoying vast information resources, users are immersed in their own "one-third of an acre" [17], confined in "information islands." The problem of the "information cocoon" is becoming increasingly severe.
Information cocoons "assist" users in avoiding the mainstream ideology, largely encroaching on the social foundation upon which the mainstream ideology leads social thought and mentality. Comprehensiveness, rigor, authenticity, and authority are the prominent characteristics of the mainstream ideology. However, under the influence of recommendation algorithms, the grand narrative style has actually become a weakness in its wide dissemination; rigor leads to a clinical decrease in attraction, comprehensiveness (a necessary condition for authenticity) causes a weakening of targeted appeal, and authority triggers a loss of personal affinity. Under the label of "personalization," users are trapped by "filtering out heterogeneous information and indulging in homogeneous information." The stacking of fragmented information further breeds user dependence on recommendation algorithms and weakens their holistic thinking. This causes users to step-by-step lose the habit and ability to actively acquire information and think independently, losing the patience to "be concerned about every matter—be it family, state, or world affairs." [18] Consequently, they cannot produce a systematic and comprehensive cognition of the mainstream ideology, making it even harder to reach identification on an emotional and value level, which invisibly raises the difficulty level of mainstream ideological dissemination.
Information cocoons also hinder the formation of social consensus at the level of mainstream ideology and generate a fertile environment for online "stratification" [19] or even group polarization. While personalized recommendations assist users in finding partners with similar interests, they also accelerate the speed of "stratification." As independent individuals, users hope to seek groups with the same interests and inclinations in cyberspace to satisfy their sense of value belonging and emotional identification. As a result, the public cyberspace is divided into various circles that are incompatible with each other. Users within these circles only pay attention to information within their own "circle" and ignore or oppose all information "outside the circle." Stuck in the "information cocoon" and unable to extricate themselves, they obstruct the social construction of the mainstream ideology and foster the amplification of user cognitive limitations. This is unfavorable to the cultivation and strengthening of a sense of social community and provides opportunities for the fragmentation and polarization of social thought.
IV. Optimization Paths for Mainstream Ideology Dissemination Under the Influence of Recommendation Algorithms
General Secretary Xi Jinping has pointed out: "If we cannot pass the hurdle of the Internet, we will not be able to pass the hurdle of long-term governance." He clearly demanded: "Explore the use of artificial intelligence in news collection, production, distribution, reception, and feedback, and use mainstream value orientations to govern 'algorithms' to comprehensively improve the ability to guide public opinion." While enjoying the technological dividends, we should be even more aware of the risks and challenges carried by technology. We must adhere to cultural subjectivity and adopt a dialectical attitude, a long-term perspective, and a comprehensive strategy to go with the flow, act according to the situation, and respond actively. We must give full play to the roles of the Party in coordinating the overall situation, the mainstream media in providing demonstration and leadership, and the party and government departments in supervision and governance. We must actively guide, utilize, and govern recommendation algorithms to effectively assist in the optimization and innovation of mainstream ideological dissemination.
1. Innovating Discourse Expression, Leading Algorithms with Mainstream Ideology
The power of discourse and the power of leadership in ideology are mutually constitutive; only by firmly grasping the power of discourse in mainstream ideology can we defend the authority of its leadership. Grasping the power of discourse and enhancing its communication effectiveness are inseparable. This necessitates strengthening rational expression, intensifying emotional resonance, and using mainstream ideology to lead recommendation algorithms, ensuring that mainstream ideology does not lose its language, its voice, or its efficacy.
The degree of recognition for mainstream ideology in cyberspace depends not only on the logic and rigor of its content but also on its modes of discursive expression and accessibility. Therefore, the organic integration of rational expression and perceptual experience is particularly important. The dissemination of mainstream ideology should strengthen the systemic, comprehensive, and logical nature of theoretical explanation and promotion, changing "preaching without reasoning" into theoretical explanation and ideological exchange. This involves selecting examples that people enjoy [20], using vivid expressions and easy-to-understand language to imbue mainstream ideological content therein, and competing for attention resources. We must utilize virtual reality and artificial intelligence technologies to achieve the transformation of mainstream ideological dissemination into tiered and three-dimensional scenes, helping users experience the profundity and wisdom inherent in mainstream ideology within a pleasant and relaxed atmosphere. We must empower the dissemination of mainstream ideology with algorithmic technology, constructing communication formats that balance quality with traffic and fuse values with interests. By staying close to social problems and the practical demands people care about, we can effectively embed grand narratives into realistic concerns, achieving an effective enhancement of the user’s sense of identification with mainstream ideology.
The cyberspace enhanced by recommendation algorithms exhibits phenomena such as light valuation and heavy entertainment, or light sublimity and heavy pragmatism. To reshape the communication ecosystem of mainstream ideology, mainstream values must be embedded into the structural processes and operating mechanisms of recommendation algorithms to correct the value biases caused by technology. Since algorithms are created by technical personnel, we can strengthen moral influence and positive guidance for these personnel through Party and Youth League organizations. We must lead the algorithms at their source, adhering to the principle of "algorithms for good," identifying and restraining value deviations and moral loopholes hidden behind the algorithms, and building a high-level network talent team that is politically firm and technically proficient. At the same time, we must optimize gatekeeping links technically, raising the priority of mainstream ideological content, increasing the exposure of positive content, amplifying the proportion of users' positive interest points, and restricting the spread of negative information to construct a cleaner [21] communication ecosystem for the innovation and expression of mainstream ideological discourse.
2. Innovating algorithmic technology to serve the dissemination of mainstream ideology
How to harness algorithmic technology and break through the "technological black box" is the key issue in controlling recommended content and effectively enhancing the leading power of mainstream ideology within the intelligent push mechanism.
On one hand, we must optimize algorithmic technology. The root cause of the weakening dominance of mainstream ideology lies in the fact that the content screening and filtering mechanisms of recommendation algorithms are aimed at traffic and profit. This causes the information pool to be flooded with entertaining and kitsch information. Therefore, optimizing the content screening and information filtering links in the push mechanism can effectively mitigate the drawbacks of information bias and cognitive narrowing, thereby assuming the responsibility of a "gatekeeper." On the basis of satisfying basic user needs, we should revise information association rules, innovate information selection and push models, reshape the proportion of information categories in the information pool, increase the weight of mainstream ideology, classify information quality levels, increase the push volume of high-level information, and eliminate low-level information to effectively improve the communication precision of the push mechanism. We must persist in the unity of instrumental rationality and value rationality, injecting elements of "emotion" and "meaning" into recommendation algorithms. By converting positive emotions and mainstream values into data models for algorithm training, we can move people with emotion, touch them with heart, and transform them with righteousness, effectively enhancing the cohesion and leading power of mainstream ideology. Simultaneously, we must embrace new technologies with an open mind, improve technological sensitivity, and achieve the innovative fusion of recommendation algorithms with technologies such as cloud computing and the Internet of Things (IoT) to broaden the channels through which technology serves mainstream ideological dissemination.
On the other hand, we must improve the transparency of recommendation algorithms. We should advocate for open-source thinking and encourage platforms to face society with sincerity, demonstrating their operating principles and risk assessment results to users. This will improve the transparency of algorithmic data, operating mechanisms, and backgrounds, safeguarding users' basic right to know and right to choose. We should add public opinion monitoring mechanisms and push-effect feedback loops, adhering to the principle of algorithmic explainability to analyze the development trends of harmful public opinion and dangerous ideological trends. We must eliminate risk sprouts at the source and provide protection and remedial measures for situations that have already caused harm. Beyond this, we should establish preference detection systems and anti-recommendation mechanisms; once user cognitive stagnation is detected, the platform should automatically push a set amount of content the user "dislikes" to help them break through "information cocoons." This will enrich the types and range of content pushed by recommendation algorithms, increase the breadth of information reception and divergent thinking, effectively resolve the issue of mainstream ideological dissemination space being squeezed, and achieve the organic unity of personalized, diversified, and mainstreamed information pushing.
3. Improving prevention systems to regulate the operation and development of algorithms
General Secretary Xi Jinping emphasized: "We must regard governing the web according to the law as a fundamental method, continue to accelerate the formulation and improvement of laws and regulations in the Internet field, and promote the management, operation, and use of the web according to the law to ensure that the Internet operates healthily on a rule-of-law track." Network society is not a place outside the law [22]. We must improve relevant rules and systems, strengthen supervision and screening, and cover all links of information filtering, pushing, and feedback to escort the dissemination of mainstream ideology in cyberspace.
Sounding and improving cyber security laws and regulations is the foundation for regulating algorithmic operations. Based on the characteristics of information dissemination and the operating mechanisms and development laws of recommendation algorithms, we should clarify the technical ethics of algorithms, divide their operating fields and jurisdictional scopes, improve industry norms and codes of conduct, and issue management opinions and policy measures. We must analyze and judge situations that may arise but are not yet legally defined, clarify which algorithm practices are legal, use the compulsory force of the law to restrain chaos in cyberspace, and prevent possible dangers. Meanwhile, legislation should be scientific and timely, closely following the characteristics of information dissemination—such as timeliness and interactivity—and following the development laws of recommendation algorithms. While maintaining their vitality, we should effectively achieve the organic unity of positive algorithmic development and the security guarantees of network society, providing a clean communication space for mainstream ideology.
Supervision and review by relevant departments are important guarantees for mainstream ideological dissemination. We must strengthen the supervision of cyber-ideology, increase monitoring and intervention in algorithmic technology and the information market, and severely punish behavior that violates laws and regulations or harms public interest, using these as reference cases for public awareness to promote the rule of law in network society. Supervision and review work must not be superficial or merely ex post facto; we should use algorithmic technology to periodically screen and audit relevant websites, conduct real-time monitoring of various platforms, dare to struggle, and "draw the sword" [23] against dangerous trends and harmful ideas to nip danger in the bud and purify the dissemination ecosystem of mainstream ideology.
4. Improving algorithmic literacy to learn to use recommendation algorithms well
To achieve effective communication between mainstream ideological content and users, we must improve the public’s algorithmic literacy and avoid the risks brought by recommendation algorithms. So-called "algorithmic literacy" refers to users achieving a basic understanding of the operating mechanisms and functions of recommendation algorithms and understanding what kinds of risks and challenges accompany them.
Algorithmic principles are the "knocking brick" [24] for cultivating algorithmic literacy; only by entering the door of recommendation algorithms can one analyze and judge what risks they will trigger. We must mobilize social resources, strengthen the responsibility consciousness of platforms, and encourage platforms to appropriately disclose source code, operating methods, processes, and R&D backgrounds. This allows users to choose and consume information services with a cautious attitude based on a preliminary understanding of recommendation algorithms. Furthermore, we should invite experts from the computer and network industries to conduct online courses and lectures on the risks caused by recommendation algorithms, filling the gaps in users' knowledge of principles and strengthening their awareness of how effects like "information cocoons," "group polarization," and "post-truth" affect harmonious social development and social cohesion.
In the era of the information explosion, recommendation algorithms are an indispensable "dharma instrument" [25]. Since the macro-environment cannot be changed, the best strategy is to transform our thinking and actively respond to risks and challenges. We should guide the public to treat recommendation algorithms with critical thinking, clarify their limitations, gain insight into their deceptiveness, consciously seek out different types of information, and expand the breadth and depth of their information needs, using changes in their own concepts and behaviors to break through the cage of algorithmic technology. Simultaneously, improving users' ability to discern and integrate information is the "silk bag strategy" [26] for risk resolution. We must use a broad knowledge base to puncture "information cocoons," use diverse information categories to dismantle "group polarization," use timely and comprehensive responses to social hotspots to offset "post-truth," and rely on systemic and comprehensive thinking to integrate "fragmented" information. By identifying and filtering negative information and resisting and rejecting harmful or vulgar content, we can optimize the preference pools of the audience for mainstream ideology, increase communication efficacy, open new frontiers for dissemination, and create a new situation for the innovative integration of recommendation algorithms and mainstream ideology.
The report to the 20th National Congress of the CPC pointed out: "We will improve the comprehensive system for network management and promote the formation of a good network ecosystem." In the era of intelligent media, where "everything is media and humans and machines coexist," recommendation algorithms are deeply embedded in social life in an "unannounced" manner. As their influence and radiation range expand daily, they constantly drive profound changes in information distribution and transmission forms, opening a new era of intelligent information dissemination. Against the backdrop of a complicated international situation and the unpredictable field of cyberspace, original information flow rules and social power structures are being broken. Mainstream ideology is easily drawn into the "data vortex" regulated by recommendation algorithms, necessitating the construction of a new order for information dissemination. We should actively embrace change, clarify the essence and characteristics of algorithmic technology, explore the fit and development trends between algorithms and mainstream ideological dissemination, and actively seek practical strategies for understanding, integrating, and harnessing algorithms. We must respond to changes scientifically, turning the "greatest variable" into the greatest increment, seizing the opportunity to occupy the main battlefield of communication in cyberspace, strengthening the awareness of ideological security among all network subjects, improving technical support and infrastructure, forcing recommendation algorithms to serve mainstream ideology, and actively preventing ideological risks induced by algorithms to further consolidate the guiding position of Marxism in the ideological field.