Next Article in Journal
New Exact Solutions of the Generalized Benjamin–Bona–Mahony Equation
Previous Article in Journal
Some Fixed Point Results on Rectangular Metric-Like Spaces Endowed with a Graph
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Journalism Model Based on Blockchain with Sharing Space

School of IT Engineering, Sookmyung Women’s University, Seoul 04310, Korea
*
Author to whom correspondence should be addressed.
Symmetry 2019, 11(1), 19; https://doi.org/10.3390/sym11010019
Submission received: 18 November 2018 / Revised: 11 December 2018 / Accepted: 17 December 2018 / Published: 27 December 2018

Abstract

:
The challenge that journalism is facing these days in the Internet mobile environment is greater than ever before. Journalism is losing its revenue structure to platform operators favoring a certain markets, and also the trust of its readers in light of fake news and infected news. To alleviate this situation, we propose a blockchain technology that is applicable to journalism in order to achieve decentralization as a reasonable alternative. The journalism model based on hybrid blockchain aims to achieve the following: the delivery of articles with sharing value, what we call proof of sharing; the distribution of roles of personalized agenda settings; and finally, the use of agora to collect public opinions. With all these, we attempt to resolve the issues with current journalism with our proposed model based on blockchain.

1. Introduction

With each communications revolution, journalists have always needed to adjust to learning new technologies in order to communicate accurate information to mass audiences. When the radio was invented, journalists mastered the skill of speaking to a mass audience in real time, and when television was invented, they learned to effectively communicate information to a mass audience through video. However, the development of the Internet completely transformed the way people worldwide are connected and destroyed the concept of “the masses.” With the development of the Internet and mobile technologies, it is no longer possible for journalists to survive by simply learning the art of technology.
To see why, consider the profit structure of traditional journalism. In short, journalists served as gatekeepers, connecting the world and the masses. The public had to pay for subscriptions to gain access to journalistic media, e.g., read newspapers or magazines or watch TV. Companies had to advertise their products by renting space in a given journalistic medium. The journalistic media monopolized or oligopolized the channel connecting the world and the public, which made it easy to earn money [1]. However, the Internet has given individuals a window to connect directly with the world. Google and Facebook have given users much more options, and consumers have adapted easily to a platform that can tailor to their tastes. Journalists, as one of these options, could not help but notice the unreasonable high advertising costs assigned to the platform operators.
Fake news and phishing articles not only create public opinions that cloud people’s objective judgment of politics, but can also arouse public distrust of journalism [2,3]. Fake news and phishing articles appear and spread more frequently on the Internet and in mobile environments [4]. Of course, going back to the birth of news, the delivery of true information has often been accompanied by political instigation and gossip. Since it is easier to earn advertising revenue with just one click, it may be argued that there is no reason why we should not manipulate the news. However, the value of journalism, i.e., its credibility, is at the heart of what makes journalism a sustainable industry in this society. Journalists have persuaded people that they are the ones who regulate power based on accuracy and speed, and who act as the defenders of civil liberties, thereby creating the reason for their existence. However, fake news and biased articles have reduced the sustainability of the industry by overlooking the trust of citizens.
Sometimes, disruptive innovation, which completely breaks down the existing industrial structure and creates a new structure, can have more influence than sustaining innovation, which complements the disadvantages but maintains the existing structure [5]. The journalism industry, which is rapidly collapsing, needs new structures and markets. Innovation in the emerging blockchain technology opens up new market possibilities in many areas. This innovation is not only sufficiently applicable to the field of journalism, it can also improve the problems discussed above. In this paper, we propose a journalism model that provides completely personalized news based on a distributed system by moving away from centralized control.
Personalized journalism based on blockchain aims to accomplish three things: First, it delivers articles with value to share. It prevents article abuse, filters out fake news, and produces more useful news. Second, it automates the process of setting the agenda. It aims to minimize the power held by both traditional journalism and new platform operators. Finally, it creates a public area, the agora (an ancient Greek term for a public open space used for assemblies and markets), where opinions are collected. By achieving these goals, critical voices can monitor power and contribute to social development.
This paper is developed as follows: Section 2 introduces related work. Section 3 describes the design of the overall architecture and hybrid blockchain with journalism. Section 4 describes more specific models. The proof-of-sharing concept is presented and personalized journalism and the agora are mentioned. Section 5 presents a scenario for the flow from proof of sharing to personalized systems. Section 6 concludes the paper.

2. Related Work

2.1. Different Directions of Journalism in Response to Technological Innovation

Attempts have been made to develop journalism in line with technological innovation. On the other hand, journalists who want to persuade the public by being more faithful to the function of journalism have been interested in paid content, such as investigative journalism. Instead of relying on advertisers, they want to break through the current crisis with competitive content [6]. Investigative journalism is aimed at consumers who are tired of the rampant soft news and are looking for professional content. Investigative journalists have sought profit models to receive subscriptions for consumers who are willing to pay for content [7,8]. However, it is already known that investigative journalism cannot completely rescue the crisis of journalism. Despite the debate on investigative journalism since the beginning of the crisis in the newspaper market, it has only become a genre of journalism and the crisis has persisted [9,10].
On the other hand, journalism has also developed by using technology in the process of creating news content. Computer-assisted reporting (CAR), introduced by Quinn in 1999 [11], and computational journalism, introduced 12 years later [12,13], are typical examples. Both are types of journalism that uses computer technology. Computational journalism differs from CAR in that it utilizes data and information processing that cannot be handled physically, supported by big data algorithms. Robot journalism, a more advanced form, further minimizes human intervention. Both computational journalism and robot journalism can be categorized as journalism based on algorithms. However, it is possible for robot journalism to make its own decisions without human control, as compared with computer journalism. Robot journalism is an automated process of searching for data and writing articles by itself according to algorithms [14,15,16]. Robot journalism is just the beginning, so it is too early to predict what impact it will have on the overall journalism industry. Nonetheless, several studies have demonstrated the feasibility of algorithm-based and automated journalism. For example, there is research on the ability to recommend articles automatically, and research showing that people are more receptive to sports articles created through algorithms [17,18,19].

2.2. Blockchain Technology

Satoshi Nakamoto first introduced the blockchain in a Bitcoin paper in 2008 [20]. A blockchain is a large, distributed public transaction ledger. This technique does not require a central authority that exists to secure trust between parties. It is the first software purposefully distributed to achieve decentralization. Previously, confirmation of the central organization was necessary for the transaction to be recognized. However, in a blockchain, all who join the blockchain recognize the transactions by recording the deal together. It is almost impossible to falsify the records of a blockchain, because such efforts would have to confront the huge computation accumulated in the meantime. Decentralization, security, and irreversibility are typical characteristics of a blockchain [21,22,23,24]. Meanwhile, Mougayar offers the following complementary definitions of a blockchain [25]:
  • In terms of technology, it is a backend database that maintains a publicly viewable distributed ledger.
  • In terms of business, it is an exchange network that can move individuals’ transactions, values, assets, etc., without intermediaries.
  • In terms of legal issues, it is a means of verifying transactions and replaces the former credit guarantee institution.
Looking at the definition, we can expect that blockchain will be fully available in other nonfinancial areas, such as content transactions, records of real estate and ownership rights, and IoT (Internet of Things) fields [26].
As the usage of blockchain diversified, the concept of the private blockchain emerged, which differs based on the characteristics of the network participants and how to derive agreements. The concept of the blockchain described above is close to the public blockchain. A public blockchain, which operates through a large number of nonspecialized participation activities, requires economic incentives such as Bitcoin and Ethereum to induce participation by nonspecific participants. On the other hand, since the private blockchain involves participation by authorized organizations, it is possible to select the algorithm for the contract according to the purpose. The hybrid blockchain, also called the consortium blockchain, is an intermediate type of private and public commodity. In the hybrid blockchain, not all participating nodes derive a consensus, but preapproved nodes derive consensus. Therefore, if the subject participating in the transaction is heterogeneous, it is a suitable type [27,28]. These blockchain technologies are very useful in journalism. We will discuss them in more detail in the following section.

2.3. State of the Art: Blockchain Technology on Media

There are many efforts to apply blockchain to media in several ways; for example, Steem, which pursues uncensored SNS (Social Networking Service) based on blockchain; Po.et, which tries to secure ownership of digital creative assets in the Internet environment; and the blockchain-based journalism platform Civil. This section introduces Steem, Po.et, and Civil (including the ultimate purpose of this paper, blockchain for journalism). This is because the goal is not just to share news, but to play a role as a media platform. The “Steem power” rule proposed by Steem to guarantee the quality of the content was used to create rules for distributing valuable articles in proof of sharing. Furthermore, the proof of existence of Po.et is in the same context as proving the identity of the copyright holder who wrote the article.

2.3.1. Steem

Steem is a distributed service that pursues blockchain-based SNS systems without censorship, operated on the STEEM cryptocurrency. User-generated content has created billions of dollars in value for social media companies such as Reddit, Facebook, and Twitter. Some advertising fees are paid for content with high traffic, but the media company running the platform gains most of the profits. Steem assumes that the community can become more active by rewarding those who contribute to the growth of the platform. In other words, Steem is a direct reward for people who make platforms active, by scoring their contributions. When people are recognized for their meaningful contributions, they keep contributing and the community grows. The challenge is how to assess the quality and quantity of participation of a large number of users and provide corresponding compensation. This requires a scalable voting process. In particular, it is necessary for the authority to allocate funds to be as distributed and decentralized as possible. What is noteworthy is that the authority to allocate funds is not completely equally distributed. The greater the contribution to the platform, the greater the voting power. However, as the voting power of those who contribute more has an influence, there is a problem in that there is a tendency to exchange votes with those who have strong Steem power instead of those who generate high-quality content [29].

2.3.2. Po.et

Po.et is a system that shares metadata and ownership information of digital creative assets through a blockchain. Currently, digital assets have no information about authors, ownership, and history, and metadata can be transformed or removed in the process of being optimized or retuned for the network. Po.et is intended for marking ownership of digital assets and the utility of use, and tracking original assets. Proof of existence is used to prove data ownership, to document time stamps, and to verify the integrity of documents without revealing actual data. Because the user proves the data ownership that generates the hash with the password, ownership conflicts can occur, and it is necessary to prove ownership of data when ownership conflict occurs. The timestamp ensures that a set of data existed at a certain time. This allows users to prove that they own the data at a certain time. In addition, proof of existence makes it easy to verify the integrity of the document, because it notifies the system of any other hash due to modifications in the document [30].

2.3.3. Civil

Civil, the journalism blockchain platform, is an alternative journalism model for existing journalism using the Ethereum token-based encryption currency called CVL. Civil is most similar to the blockchain media platform that this paper aims to describe. Civil is an open market that allows journalists and readers to directly trade news using blockchain technology. As the journalists and readers deal directly with each other, they intend to be free from the influence of advertisers, political pressure, all kinds of censorship, and portals that select articles. The Civil newsroom operates as a decentralized self-governing organization, which has five types of participants:
  • Journalism advisory board: An independent organization consisting of journalism experts that functions as an adjustment in case of conflict in the Civil network.
  • Officers: Manage the newsroom in accordance with charters, which are created with the approval of newsmakers and citizens. They are responsible for newsroom operations.
  • Newsmakers: All journalists, including photographers and video reporters, editors, illustrators, and others who create newsroom content.
  • Citizens: News consumers. Citizen can buy access to articles with CVL tokens issued by Civil.
  • Fact checkers: Check the basics of journalism: facts. Fact checkers work to earn rewards for their CVL tokens and to gain a high reputation within the network.
The newsroom based on the above participants is driven by what citizens want to know. Citizens gather CVL tokens to support reporting on specifics topics, and newsmakers gather to cover it. This market not only develops organically in response to demand, but also enables healthy competition among journalists [31,32].
Civil, a promising journalism platform for the future, is currently under development and is not yet fully embodied. However, Civil should reconsider the assumption that citizens will voluntarily purchase CVL tokens and supports journalists. In fact, only 5% of people subscribe to digital news for a fee, according to a report that analyzed digital news consumption in 36 countries worldwide in 2017; 54% of respondents said that they did not have subscriptions because they could watch the news for free [33]. By the way, Civil ruled out any possibility of advertising in order to have transparent journalism free from outside pressure. Even when newspapers were very influential, subscription fees were not their only source of revenue. Delivering information through advertising does not only function negatively. Delivering real estate ads, job placement ads, and descriptions of new products that consumers do not know about is necessary for both companies and consumers. In addition, writing articles relying on voluntary support from citizens can cause overcrowding on popular topics. The problem is that overcrowding on a particular subject can undermine the diversity of articles. Nonetheless, Civil’s problem consciousness is almost in agreement with the problem consciousness of our paper. In particular, the argument that the distributed system should share the power of the media is entirely in agreement. The next section will examine the changing patterns and problems of the journalistic ecosystem and discuss the possibilities and positive impacts of distributed systems.

3. Overview of the Proposed Model

3.1. Centralized Sand Distributed Systems for Journalism

Figure 1 shows a schematic of the journalism ecosystem before the development of the online environment. When a small number of publication houses monopolized the journalism market, their strength was absolute. The press had authority over setting agendas, and readers had no choice but to receive the news that the press decided was important for them. At that time, the press called readers “the masses” and treated them as one mass. A mass by definition is not able to choose the news according to personal wishes. Despite the emergence of new media companies, they can consistently monopolize the market because of the unique characteristics of media products. Media products are subject to network externality. After the initial entry, the press secured some readers, and the media gained readers by being heavily advertised and capitalized. Due to quality articles produced by the capital-rich press, readers were gained again. This made it difficult for media companies that entered later to survive [34]. For example, in certain regions of the United States, there is now only one local newspaper left. The exclusive press was afraid of business, and even the government. When media attention turned to them, the companies were at a disadvantage. Under these circumstances, the press had no opponents to check its operations, only itself. In other words, they could manipulate the information as much as they liked, or lead public opinion in the direction they wanted [35].
However, as the Internet developed, the situation gradually reversed. Figure 2 shows a schematic of the relationship between the newly emerging platform operators and the heavily competitive press.
People can access Internet platforms anytime and anywhere, and have more choices. Advertisers no longer need to meet consumers only through spaces provided by the press. Mobile developments have accelerated this phenomenon. As people spent more and more time on new media, they became more likely to receive articles on the Internet and mobile platforms. Subscription fees through traditional methods have dropped drastically. In addition, Internet media companies or individual content producers have entered the market, further competing for limited advertising expenditure. In this environment, it was a stimulus that the press chose to gain more attention. They wrote increasingly provocative articles to receive even one more click, and they did not hesitate to write fake news and phishing articles. Fake news spread quickly and easily on the high-speed Internet network. At the same time, the public’s trust in the press dropped. Another problem is that a small number of platform operators have become too strong. Although it looked like an individual choice, article layouts and recommended algorithms were tightly locked to the secrecy of the platform operators. There was also a problem with privacy protection.
The traditional press had the power to manipulate public opinion. The present press prints a large volume of inferior articles swayed by platform operators. Both have led to a crumbling of public confidence in the press. The common point of both problems is that centralized power has the editing rights. Blockchain opens up the possibility of a decentralized network. We propose that if centralized editing fails in journalism, the solution could be to distribute editing rights using blockchain technology. Figure 3 shows a schematic of the ecosystem of journalism on a blockchain. In distributed journalism, readers and journalists can meet directly. In addition, readers own the data for all articles generated, and they can receive news through automated recommendation algorithms.

3.2. Journalism on Hybrid Blockchain

In journalism on blockchain, the data stored in the chain are articles, thus it is a database composed of linked articles. The reason why we utilized hybrid blockchain, not public blockchain or private blockchain, is that journalists and consumers who participate in the network have different properties. The hybrid blockchain is appropriate when there are heterogeneous participants, as we have seen above. The process of verifying facts and generating information is not something that anyone can do, like financial transactions, and it is professional. For this reason, participants allowed to perform the work should have prior approval. At the same time, the created article blockchain should be accessible to anyone participating in the network. In other words, both journalists and consumers have the right to read the article blockchain, while only journalists have the right to write articles.
It is possible, with the consensus of preapproved participants or journalists, to create new articles on a journalism hybrid blockchain. This process of consensus is different from the mining activities, or proof of work, of the public blockchain, because journalism has different characteristics from the general transaction process. An article includes a viewpoint of the reporter looking at the facts. Putting a reporter’s viewpoint in the article does not mean that the article is false, it just conveys one aspect of the facts. Journalists can write about various aspects of the same fact. Therefore, we should not censor article writing by prior consent. On the other hand, not all information is worth sharing with people. For example, the news that my dog died may not be of any benefit to other people. The proof of sharing process is the process of authenticating whether or not a share is worth writing an article.

4. Proof of Sharing, Self-Regulating, Personalization, and the Agora

4.1. Proof of Sharing

Proof of sharing is the process of getting certified by preapproved journalists to link to an existing article chain when an article is created. Figure 4 shows a schematic of the simple process of proof of sharing.
Journalist A creates a new article. The article goes through a redundant verification procedure for journalist B. This is the process of preventing similar articles from appearing many times. Journalist B checks whether the article is truly new. In order for journalists and consumers to have different access rights to the article blockchain, and to function as a journalist, the node should undergo a reliability verification process. This process assumes trust in journalist B. Duplicate verification consists of simple comparisons between new and existing articles. This does not require any subjective judgment. Therefore, a single randomly extracted journalist with an automatic duplicate check can sufficiently judge it. Since all journalists will also verify the value of sharing after duplicate screening operations, this has to save time and shorten the process of article creation.
All journalists will be able to receive confirmed articles immediately. Journalists who receive an article may leave brief comments, which may be positive or negative, or not comment at all. The more comments that are left in a limited amount of time, the more valuable the article is, and the more it is worth sharing. The time for commenting should be limited to ensure timeliness.
Comments are stored together in a block of the article that cannot be modified, but only added to by authenticated users. At this time, the criteria for judging what it is worth sharing are determined by the rules developed when the article blockchain was created. This is a consensual rule, not a decision made by a particular leader or delegate. Therefore, network participants can change the rules by consensus. In a more sophisticated model, different criteria may apply, depending on whether the article is breaking news or not. As mentioned earlier, the right to write to the article blockchain is in the hands of trusted journalists. Finally, journalists link articles that they think are worth sharing to an existing article blockchain.

4.2. Self-Regulating Rule

This section discusses the attacker who can reduce the platform’s reliability and proposes a rule for solving the problem. Even if an article passes a comment rule (which, according to proof of sharing, can only be stored in an article chain when a certain number of comments is received), the article may still be fake news or undesirable content. If fake news continues to circulate through comment rules, the platform may be unreliable. In order to prevent this, the platform should continue to validate journalists.
We already ask journalists to go through the verification process through comment rules themselves before an article is distributed. However, the comment rules only determine whether the article is worth sharing and distributes articles that are of interest to the majority of readers. The reason why we cannot completely exclude fake news before it is distributed is freedom of expression. We can never give clear criteria for what fake news is. To judge the veracity of an article, therefore, it is preferable to use follow-up measures rather than advance actions.
The journalism model based on blockchain includes the forced exit rule, whereby journalists who debase the platform by continuously distributing fake news must leave. Journalists who may be stakeholders should not proceed with this verification. This is because if a group with the same interests evaluate it, objectivity may deteriorate. In order words, groups that grant journalists status and groups that deprive journalists should be different. In Section 3.2, we state that granting journalist status is possible by consensus among journalists participating in the network. Then, the qualification to deprive journalists of their status should be by consumers who become exclusive groups with journalists. For example, if consumer judge that an article is fake news, they may warn the journalist who wrote the article. Journalists who receive these warnings across multiple articles will automatically no longer be able to write articles. The self-regulating rule cannot block all journalists who distribute bad quality articles, but it can be at least a supplement to the ecosystem to self-regulate itself.

4.3. Personalized Journalism

The following discusses the matter of article distribution. This is the agenda setting, and it is the ultimate reason why the press of the platform holds such great power. Until now, organizations that published articles had difficulty knowing all their readers’ personal information, so they delivered common articles to everyone, or provided articles targeting groups whose characteristics could be predicted. Personalized journalism provides everyone with the news they really need. It becomes possible for every individual to access the distributed article blockchain. The process of making personalized journalism through distributed article blockchain is as follows:
Any articles contained in an article blockchain are likely to be listed for the participants. All participants can view all the articles in the blockchain. However, the final set of articles is selected individually according to the digital identities of the network consumers. The provider of such a service is a journalist. Personalized recommendation systems will run almost automatically, unlike platforms or media outlets that have authority over all the agenda settings. This is because the above process is completely independent of agenda settings. Given that a third party cannot control all journalists participating in a blockchain and cannot manipulate all the individual agenda settings, it is virtually impossible to control the media and manipulate public opinion in personalized journalism.
Furthermore, the personalized journalism model is also meaningful as a business model. Until now, the media’s main source of revenue has been advertising. Personalized recommendation systems are highly attractive to advertisers that want to increase the effectiveness of their advertising by targeting suitable consumers. Paid services that receive content fees in exchange for providing articles to consumers can be a business model for journalism, but may also offer free services along with advertisements through a personalized recommendation system.

4.4. The Agora

The agora was a public space where citizens voiced their opinions in ancient Athens. The role that citizens expect of the media is not just information delivery, but also criticism by the intelligence that authorities are leading citizens in the wrong direction. They must construct the agora of public opinion and make the citizens’ handpicked power fear their voices. The agora organically connects with the preceding article blockchain, but it is obviously a different space. The public space for the agora can be created as a public blockchain, and the data stored in the agora blockchain are the arguments of participants. Anyone can participate, and read and write freely. Even robots can form a node. A robot certified as a journalist in the article blockchain may function as a journalist collecting public opinion from the blockchain and generate new articles. The detailed structure of the agora blockchain is outside the scope of this paper, but we will reinforce it in later studies.

5. Scenario

5.1. Scenario

We discussed the process of creating articles through proof of sharing and the possibility of personalized distribution. In this section, we look specifically at various situations that can arise in the process of proof of sharing and personalized recommendation systems.
Figure 5 is a schematic of how to store articles in an article blockchain and deliver them to customers. It is based on proof of sharing and a personalized recommendation system, but it does not give an absolute direction. Participants can change the rules by consensus, further developing the system.
In Figure 5, the red shapes depict the process of proof of sharing and diagram a flow to store articles in an article blockchain. This makes sure that articles do not duplicate existing articles before checking their timeliness. Breaking news articles should be stored and be immediately visible to consumers. As the articles do not go through the process of verifying shared values, a solid standard is required that can be categorized as a breaking news item. Even if it is not breaking news, the timeliness of the article is still important. Time-setting is the process of limiting the time within which journalists can leave comments. Any article that does not have sufficient comments within that time is discarded. The next step is to judge whether the amount of comments is sufficient or not. The example above stores articles that have more than the average number of comments. Journalism model participants decide by consensus how to determine the storage criteria. For example, one criterion can be an average number of comments.
Two assumptions are necessary for the proof of sharing mechanism to work smoothly. First, there is no immediate reward for the tasks required for the process, i.e., confirming duplication and determining whether an article is worth sharing. Second, these tasks are essential to the formation of the platform, and participating journalists should be mandated to accomplish them. For example, journalists should immediately review whether their articles duplicate other articles, and they should make comments on whether there is shared value for any articles. Based on the current situation, where an average of 6.42 articles are posted on the Internet per minute (1,109,028 articles published from January to April 2018 [36]), if six journalists each comment on an article, then the probability that the article will get a comment is 0.16. As another example, suppose that 120 journalists each produce two articles in a day. When articles produced in a day are distributed evenly over time, 10 articles per hour are issued. Within a one-hour time frame and comments on two articles by each journalist, each article receive 20% of the comments, on average, as shown in the above assumption. If a journalist comments on one article each time he or she creates an article, the article gets an average of 10% comments. As such, the frequency of occurrence of articles is closely related to the number of comments. However, apart from this, journalists must reach a consensus on how many articles to evaluate each time they create an article. Even if an article receives an average of 20% of comments, journalists must agree whether the average value is the basis for storing articles in the blockchain.
The green diamonds in Figure 5 describe the flow process of personalized delivery. Personalization is performed in many ways, based on the digital identities of consumers. Consumers can selectively display information about themselves when they access the article blockchain. For example, consumers A to D provide examples of criteria that consumers can choose based on the characteristics of the article blockchain, while consumer E shows personalization in terms of the content of articles. Personalized service for consumer E becomes possible when the classification according to the content of the article is included as an element of the article block. This model requires further refinement. Consumer A wants to see articles that are more favorable to journalists. No matter how many times the article receives comments, when there are more negative comments than positive comments, consumer A cannot receive the article. Consumer B wants articles that receive more attention than what the criteria in the blockchain allow. Consumer B cannot receive articles that do not have more than half the number of comments. On the other hand, consumers who do not present any information, such as consumer C, can see all saved articles. However, as information about the consumer’s preferences accumulates, the model will recommend articles based on information about features of preferred content or other consumers’ preferences with similar choice. Consumer D wants opinion-activated articles. If comments show the significance of the article as determined by journalists, the agora is the importance of the article as judged by the public. In addition, there could be a variety of consumer preferences.
All the figures presented in the scenario above and the information about consumers A to E are assumptions. These assumptions can fluctuate over time. However, by providing several scenarios, we can see the effectiveness of journalistic models when they are on the blockchain. First, only articles that have undergone duplication testing can start the journalism blockchain storage process. Second, the editorial right to the news is distributed to all journalists in the blockchain. Journalists can decide whether to comment on an article or not, and they may agree on the standard for the value of sharing, e.g., the average in the above example. In other words, no journalist has absolute authority over the platform. Third, consumers read the article blockchain through different windows. Each consumer has a different arrangement of articles, which is subject to various digital identities, so no third party can control them.

5.2. Experimental Setup and Simulation

As we discussed in Section 5.1, receiving comments is the most important criterion for article distribution. In this section, we provide an analysis that can be referred to when setting the criteria for linking an article to an article chain (in the example in the previous section, receiving an average number of comments or more was the criterion). In the preceding scenarios, judgments about whether an article is commented on and distributed are influenced by three variables: the number of published articles per minute, the number of comments per article, and time setting. The number of published articles per minute refers to the assumption about the average number of articles per minute. The number of comments per article is the number of comments that journalists must leave in order to publish articles. Time setting refers to the maximum amount of time that comments can accumulate before determining whether or not to issue an article. Only articles that obtain the number of comments that meet the rule after the set time are automatically distributed. When a journalist publishes an article, the time setting delays the time until the model distributes it to consumers and sequentially attaches it to an existing node in chronological order. This solves the problem of the validity of node replication in existing blockchains. The model distributes articles published first, and if there are chains with different time sequences, the model replicates chains in sequential order.
Table 1 is a comparison of six cases that vary depending on the number of articles issued per minute, time setting, and the number of comments needed to publish an article. We set three variables through algorithms and randomly extracted articles and required that comments be made. We set minimum and maximum values for the number of articles issued per minute based on the actual average number of articles published per minute per day [36]. Time setting also considered the way in which actual articles were published and circulated. Based on the timeliness, we considered that 24 h would be the maximum. Finally, the number of comments required per article was set to at least two. This is because split-ticket voting can provide voters with an election insurance policy rather than a straight vote [37]. Especially since it is not a vote to select the best articles, it should be possible to vote not only for articles favorable to voters but also for what they believe is necessary. Table 1 also contains information about how many articles can be published in a year. The table gives us an idea of how many comments each article will receive, roughly, for the entire year. Histograms below show the distribution of articles according to the number of comments (Figure 6). The x-axis of the graph represents the number of comments, and the y-axis represents the number of articles that received comments.
As a result, the average of the number of comments on an article converges to the number of comments required when a journalist publishes an article. Second, as the number of comments increases, the number of articles about the number of comments becomes a normal distribution. Third, the average number of articles published or the time setting does not affect the number of articles for the number of comments. However, we should be careful that as the number of published articles increases or the time frame becomes longer, the number of articles journalists should consider commenting on also increases. Finally, when we store an article based on the set number of comments, that is, the average number of comments an article can receive, we know that the model stores only about 60% of the articles published by journalists. If you want to increase or decrease the number of articles stored in the chain, you can adjust the blockchain under the agreement of the participants.

6. Conclusions

In this paper, we proposed a journalism model based on blockchain with sharing spaces. It is based on a society of common ownership with distributed ledger technology. The design of the model reflects the unique characteristics of journalism. First, we applied a hybrid blockchain to the journalism model. This is because the access rights of journalists and consumers involved in the blockchain network must be different. Preapproved journalists have both write and access rights to the blockchain. Consumers cannot create articles, but have access to the chain created by journalists. Second, we introduced the concept of proof of sharing, which is a kind of verification process for linking newly created articles to existing chains, and is conducted by preapproved journalists. Since journalists can write from multiple perspectives on the same fact, it is meaningless to verify the creation with a 51% rule. In this process, journalists verify that an article is worth sharing. Third, we proposed a rule that enables self-regulating action for blockchain journalism. The space where news is formed is inevitably threatened by false information. The self-regulating rule helps the platform’s self-regulatory behavior by allowing different groups to manage the inflow and exit of journalists, who are the main agents for publishing articles on the platform. Fourth, personalized journalism is a function that recommends articles to individuals in personal places with a distributed article chain. As the agenda is set up individually, it has the effect of dispersing the great power of the traditional media. Finally, the argument tag, attached separately from the article creation, serves as a sort of agora for people to exchange opinions. Through these functions, we hope to use it as a new, alternative journalism to solve the chronic problems of journalism today.
In this paper, we did not mention any incentive for journalists. There are two major areas where journalists can receive incentives: the cost of evaluating articles during the proof-of-sharing process and the cost of content they write. Until now, journalists have been paid to write articles by media companies or platform operators. Their role was to write articles. However, journalists gradually have to become the owners of the platform themselves. The reason journalists lose power in the media market is because the space where people gather is taken away by platform operators. Platform operators simply bring content together. Imagine a social networking site like Facebook or a portal site like Naver. They just collect the content in one space and then collect people in that space. When people gather there, businesses that want to sell information to people pay attention to it. Where many people gather, more people and businesses are concentrated. This is called network externality. It is hard to see why people gather on the platform simply because of the content. The reason why platform operators, not content creators, are able to earn more revenue is that they spent their time and effort building the space where people gather [38]. However, we know that journalism has collapsed as the responsibility for the space in which people gather has been passed on to nonjournalists who are platform operators. Therefore, journalists starting in blockchain journalism are required to take responsibility for the platform at the same time. Compensation for proof of sharing is not measurable. Proof of sharing even considers not voting as a meaningful gesture. The process of proof of sharing is essential to maintain a blockchain journalism platform. If someone is given the necessary responsibility to maintain a platform, it should be a journalist who will generate revenue through the blockchain journalism platform.
In this paper, we did not discuss whether it would be better to pay journalists directly in the form of content subscription fees charged to consumers or via advertisers. That is because, although there is a direct connection between consumers and journalists, it is not certain that consumers who are already familiar with free content will pay to subscribe to the news. However, since the proposed journalism model eliminates the broker between advertisers and journalists, and it is difficult for advertisers to control all journalists participating in the network, the probability of providing biased articles to interest groups is very low. This paper describes the general structure of a blockchain-based journalism model but not the processing capacity or specific algorithms required for the network in question. We expect to develop these in subsequent studies.

Author Contributions

B.K. proposed the topic and wrote the paper; Y.Y. corrected the design and improved the quality of the manuscript.

Funding

This research was funded by the Ministry of Science and ICT, Korea, under the Information Technology Research Center support program (IITP-2018-2016-0-00311) supervised by the Institute for Information and Communications Technology Promotion.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Mc Chesney, R.W. Digital Disconnect: How Capitalism Is Turning the Internet Against Democracy; New Press: New York, NY, USA, 2013; ISBN 9781595588913. [Google Scholar]
  2. Allcott, H.; Gentzkow, M. Social media and fake news in the 2016 election. J. Econ. Perspect. 2017, 31, 211–236. [Google Scholar] [CrossRef]
  3. Mihailidis, P.; Viotty, S. Spreadable spectacle in digital culture: Civic expression, fake news, and the role of media literacies in “post-fact” society. Am. Behav. Sci. 2017, 61, 441–454. [Google Scholar] [CrossRef]
  4. Vosoughi, S.; Roy, D.; Aral, S. The spread of true and false news online. Science 2018, 359, 1146–1151. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  5. Christensen, C.M. The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail; Harvard Business School Press: Boston, MA, USA, 1997; ISBN 9780875845852. [Google Scholar]
  6. Aucoin, J.L. The Evolution of American Investigative Journalism; University of Missouri Press: Columbia, MO, USA, 2005; ISBN 9780826217462. [Google Scholar]
  7. About the ICIJ. Available online: https://www.icij.org/about/ (accessed on 26 December 2018).
  8. About Us-ProPublica. Available online: https://www.propublica.org/about/ (accessed on 26 December 2018).
  9. Park, R.E. The Natural History of the Newspaper. Am. J. Sociol. 1926, 29, 273–289. [Google Scholar] [CrossRef]
  10. Sloan, W.D.; Parcell, L.M. American Journalism: History, Principles, Practices; McFarland: Jefferson, NC, USA, 2002; pp. 209–218. ISBN 978-0786413713. [Google Scholar]
  11. Quinn, S. Computer-Assisted Reporting in Australia: Diffusion of the Internet at Daily Newspapers. Ph.D. Thesis, Graduate School of Journalism, University of Wollongong, Wollongong, Australia, 1999. [Google Scholar]
  12. Cohen, S.; Hamiltion, J.T.; Turner, R. Computational journalism. Commun. ACM 2011, 54, 66–71. [Google Scholar] [CrossRef]
  13. Weinberg, S. The Reporter’s Handbook: An Investigator’s Guide to Documents and Techniques; St. Martin’s Press: New York, NY, USA, 1996; ISBN 9780312101534. [Google Scholar]
  14. Robot Seeks Press. Available online: http://www.bloter.net/archives/232289 (accessed on 14 July 2015).
  15. Kim, D.; Kim, S. Newspaper companies’ determinants in adopting robot journalism. Technol. Forecast. Soc. Chang. 2017, 117, 184–195. [Google Scholar] [CrossRef]
  16. Kim, D.; Kim, S. Newspaper journalists’ attitudes towards robot journalism. Telemat. Inf. 2018, 35, 340–357. [Google Scholar] [CrossRef]
  17. Lindén, C.G. Algorithms for journalism: The future of news work. J. Media Innov. 2017, 4, 60–76. [Google Scholar] [CrossRef]
  18. Clerwell, C. Enter the robot journalist. Journal. Pract. 2014, 8, 519–531. [Google Scholar] [CrossRef]
  19. Kim, D.; Lee, J. Robot journalism: Algorithmic approach to automated news article generation. Korean J. Journal. Commun. Stud. 2015, 59, 64–95. [Google Scholar]
  20. Nakamoto, S. Bitcoin: A Peer-to-Peer electronic cash system. Available online: https://bitcoin.org/bitcoin.pdf (accessed on 26 December 2018).
  21. Swan, M. Blockchain thinking: The brain as a decentralized autonomous corporation. IEEE Technol. Soc. Mag. 2015, 34, 41–52. [Google Scholar] [CrossRef]
  22. Sharma, P.K.; Moon, S.Y.; Park, J.H. Block-VN: A distributed blockchain based vehicular network architecture in smart city. J. Inf. Process. Syst. 2017, 13, 184–195. [Google Scholar]
  23. Moon, S.Y.; Park, J.H. efficient hardware-based code convertor of a quantum computer. J. Converg. 2016, 7, 1–9. [Google Scholar]
  24. Finogeev, A.G.; Parygin, D.S.; Finogeev, A.A. The convergence computing model for big sensor data mining and knowledge discover. Hum.-Cent. Comput. Inf. Sci. 2017, 7, 11. [Google Scholar] [CrossRef]
  25. Mougayar, W. The Business Blockchain: Promise, Practice, and Application of the Next Internet Technology; John Wiley & Sons: Hoboken, NJ, USA, 2016; ISBN 9781119300311. [Google Scholar]
  26. Tapscott, D.; Tapscott, A. Blockchain Revolution: How the Technology Behind Bitcoin Is Changing Money, Business, and the World; Penguin Random House: New York, NY, USA, 2016; ISBN 9751101980132. [Google Scholar]
  27. Lai, R.; Chuen, D. Blockchain-from public to private. In Handbook of Block Chain, Digital Finance, and Inclusion; Chuen, D., Deng, R.H., Eds.; Academic Press: West Bath, UK, 2018; Volume 2, pp. 145–177. ISBN 9780128122822. [Google Scholar]
  28. Jang, K.J. The study on innovative financial services of business models using blockchain technology. e-Bus. Stud. 2017, 18, 113–130. [Google Scholar] [CrossRef]
  29. Steem: An Incentivized, Blockchain-Based, Public Content Platform. Available online: https://steem.io/SteemWhitePaper.pdf (accessed on 26 December 2018).
  30. About the Po.et Foundation. Available online: https://www.po.et/about (accessed on 14 February 2018).
  31. Civil: Self-Sustaining Journalism-Medium. Available online: https://medium.com/@Join_Civil/civil-self-sustaining-journalism-a5caa49005c3 (accessed on 20 June 2017).
  32. GitHub-Joincivil/Whitepaper: Introducing Free, Independent and Self-Sustaining Journalism, Enabled by Civil’s Decentralized Newsmaking Platform. Available online: https://github.com/joincivil/whitepaper (accessed on 25 October 2017).
  33. Kim, S.H.; Kim, W.G. Digital News Report 2017: Korea; Korea Press Foundation: Seoul, Korea, 2017; pp. 54–57. ISBN 978-89-5711-455-8. [Google Scholar]
  34. Gabszewicz, J.J.; Resende, J.; Sonnac, N. Media as multi-sided platforms. In Handbook on the Economics of the Media, 1st ed.; Picard, R.G., Wildman, S.S., Eds.; Edward Elgar Publishing: Cheltenham, UK, 2015; pp. 3–35. ISBN 9780857938886. [Google Scholar]
  35. Herman, E.S.; Chomsky, N. Manufacturing Consent: The Political Economy of the Mass Media, 2nd ed.; Pantheon Books: New York, NY, USA, 2002; ISBN 0375714499. [Google Scholar]
  36. BigKinds: News Bigdata & Analysis. Available online: https://www.kinds.or.kr/ (accessed on 1 May 2018).
  37. Burden, B.C.; Helmke, G. The comparative study of split-ticket voting. Elect. Stud. 2009, 28, 1–7. [Google Scholar] [CrossRef]
  38. Anand, B. The Content Trap, 1st ed.; Penguin Random House: New York, NY, USA, 2016; ISBN 9780812995381. [Google Scholar]
Figure 1. Centralized system with monopolistic press.
Figure 1. Centralized system with monopolistic press.
Symmetry 11 00019 g001
Figure 2. Centralized system with highly competitive press and platform operators.
Figure 2. Centralized system with highly competitive press and platform operators.
Symmetry 11 00019 g002
Figure 3. Distributed system by hybrid blockchain for journalism.
Figure 3. Distributed system by hybrid blockchain for journalism.
Symmetry 11 00019 g003
Figure 4. The process of proof of sharing.
Figure 4. The process of proof of sharing.
Symmetry 11 00019 g004
Figure 5. Flow diagram of article delivery.
Figure 5. Flow diagram of article delivery.
Symmetry 11 00019 g005
Figure 6. Histograms for the number of articles according to the number of comments: (a) Case 1; (b) Case 2; (c) Case 3; (d) Case 4; (e) Case 5; (f) Case 6.
Figure 6. Histograms for the number of articles according to the number of comments: (a) Case 1; (b) Case 2; (c) Case 3; (d) Case 4; (e) Case 5; (f) Case 6.
Symmetry 11 00019 g006
Table 1. Six cases based on three variables: articles, time setting, comments.
Table 1. Six cases based on three variables: articles, time setting, comments.
VariablesCase 1Case 2Case 3Case 4Case 5Case 6
Number of published articles per minute55105105
Time setting12 h24 h12 h12 h12 h12 h
Number of comments per article2225510
Total articles per unit time3,6007,2007,2003,6007,2003,600
Comments generated per unit time101020255050
Period1 year1 year1 year1 year1 year1 year
Total articles2,620,8002,613,6005,241,6002,620,8005,241,6002,620,800
Average comments per article2.00002.00002.00005.00005.000010.0000
0 Comment354,616
(13.53%)
354,017
(13.55%)
709,676
(13.54%)
17,568
(0.67%)
35,077
(0.67%)
122
(0.00%)
1 Comment709,778
(27.08%)
707,411
(27.07%)
1,418,577
(27.06%)
88,044
(3.36%)
176,695
(3.37%)
1,172
(0.04%)
2 Comments709,026
(27.05%)
707,034
(27.05%)
1,418,315
(27.06%)
221,040
(8.43%)
442,165
(8.44%)
5,867
(0.22%)
3 Comments472,705
(18.04%)
471,020
(18.02%)
946,105
(18.05%)
367,475
(14.02%)
737,274
(14.07%)
19,557
(0.75%)
4 Comments236,842
(9.04%)
236,263
(9.04%)
472,482
(9.01%)
461,097
(17.59%)
918,947
(17.53%)
49,615
(1.89%)
5 Comments94,266
(3.60%)
94,825
(3.63%)
189,514
(3.62%)
459,203
(17.52%)
918,709
(17.53%)
99,599
(3.80%)
6 Comments31,616
(1.21%)
31,322
(1.20%)
63,413
(1.21%)
383,135
(14.62%)
765,542
(14.61%)
165,040
(6.30%)
7 Comments9,063
(0.35%)
8,886
(0.34%)
17,909
(0.34%)
273,642
(10.44%)
546,814
(10.43%)
236,557
(9.03%)
8 Comments2,279
(0.09%)
2,208
(0.08%)
4,391
(0.08%)
171,452
(6.54%)
342,374
(6.53%)
295,213
(11.26%)
9 Comments478
(0.02%)
498
(0.02%)
991
(0.02%)
94,813
(3.62%)
190,805
(3.64%)
327,803
(12.51%)
10 Comments106
(0.00%)
96
(0.00%)
184
(0.00%)
47,380
(1.81%)
95,172
(1.82%)
327,603
(12.50%)
11 Comments24
(0.00%)
14
(0.00%)
34
(0.00%)
21,641
(0.83%)
43,355
(0.83%)
298,135
(11.38%)
12 Comments1
(0.00%)
5
(0.00%)
7
(0.00%)
8,923
(0.34%)
18,033
(0.34%)
248,443
(9.48%)
13 Comments0
(0.00%)
2
(0.00%)
3,510
(0.13%)
6,897
(0.13%)
190,720
(7.28%)
14 Comments1
(0.00%)
1,263
(0.05%)
2,554
(0.05%)
135,958
(5.19%)
15 Comments408
(0.02%)
855
(0.02%)
91,539
(3.49%)
16 Comments156
(0.01%)
230
(0.00%)
56,949
(2.17%)
17 Comments32
(0.00%)
83
(0.00%)
33,425
(1.28%)
18 Comments14
(0.00%)
15
(0.00%)
18,709
(0.71%)
19 Comments2
(0.00%)
1
(0.00%)
9,776
(0.37%)
20 Comments0
(0.00%)
3
(0.00%)
4,817
(0.18%)
21 Comments1
(0.00%)
2,326
(0.09%)
22 Comments1
(0.00%)
1,099
(0.04%)
23 Comments455
(0.02%)
24 Comments175
(0.01%)
25 Comments78
(0.00%)
26 Comments33
(0.00%)
27 Comments12
(0.00%)
28 Comments3
(0.00%)

Share and Cite

MDPI and ACS Style

Kim, B.; Yoon, Y. Journalism Model Based on Blockchain with Sharing Space. Symmetry 2019, 11, 19. https://doi.org/10.3390/sym11010019

AMA Style

Kim B, Yoon Y. Journalism Model Based on Blockchain with Sharing Space. Symmetry. 2019; 11(1):19. https://doi.org/10.3390/sym11010019

Chicago/Turabian Style

Kim, Byeowool, and Yongik Yoon. 2019. "Journalism Model Based on Blockchain with Sharing Space" Symmetry 11, no. 1: 19. https://doi.org/10.3390/sym11010019

APA Style

Kim, B., & Yoon, Y. (2019). Journalism Model Based on Blockchain with Sharing Space. Symmetry, 11(1), 19. https://doi.org/10.3390/sym11010019

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop