Platform regulation and “overblocking” – The NetzDG discourse in Germany

IT blog corpus

To analyze the discussion about NetzDG in different media settings, we compiled two text corpora. The first one is made up of blogs featuring news related to information technology (IT) and the discussion of laws and regulation surrounding the internet. Following research on blogs/weblogs, we define blogs as consisting of dated entries available online, often managed by a broadly available publishing tool or web space (boyd, 2006; Glance et al., 2004). The IT blog sphere (or the subsystem of blogs on IT questions) that we capture with this corpus is highly relevant for our research, since it represents an expert discourse regarding matters of technology and society that identifies important topics early on and that may have an agenda-setting function (Barbaresi and Pohlmann, 2020).

Specific tech blogs first evolved separately from, and in opposition to, traditional mass media. Amateur blogs are often characterized by their capacity to open up public space for the debate of socially relevant issues, enabling more active civic participation in the public discussion of political matters (Chu, 2019). However, the small communities initially found on blogs have largely been replaced by commercially driven websites targeting passive readers. This has certainly impacted the content of discussions. Consequently, more professional “IT news portals” may have developed and taken on the status of influential information providers, agenda setters, and gatekeepers (Barbaresi and Pohlmann, 2020). Netzpolitik.org is a notable example of a website that started as a blog or blog-like platform and has transformed over time into a hybrid between blog and IT news portal (Mey, 2010). These platforms have writers and staff continuously producing posts on tech-related topics. Due to this professionalization, they have become providers of tech and policy development news and commentary, who closely follow experts and decision-makers in the field. These IT news portals attentively monitor the discussion within the IT blog sphere, taking up topics and threads from this network of specialists. Therefore, their articles are monitored and reproduced by more traditional media outlets such as nationwide newspapers and their technology editors. We assume that some of these tech blogs and IT news portals have a recognizable impact on the discussion of tech topics within the public, eventually impacting policy discussions in political circles due to the attention that they receive from traditional news sources (Barbaresi and Pohlmann, 2020).

Corpus building

We built the IT blog corpus in cooperation with the Berlin-Brandenburg Academy of Sciences and Humanities (BBAW) by discovering relevant portions of the web semi-automatically by preselecting hundreds of sources according to their relevance to the topic. We first compiled a list of German IT blogs and websites by identifying the main websites in this field and searching for similar sites and keywords. Aside from specific IT news and policy portals (e. g., netzpolitik.org) and IT product portals (e. g., mobilegeeks.de), we collected blogs by IT lawyers and those of scholars, intellectuals, and journalists working in the fields of communication and media studies, tech, law, policy, and philosophy. After finding these initial sites, we manually extended the list by handpicking recommendations from the engine similarsites.com that fit the corpus (Barbaresi and Pohlmann, 2021).

The detected relevant websites were then retrieved starting from their homepage. Potential pages were identified either by using available sitemaps or by applying web-crawling techniques to identify the links. All the content that could be found was archived. Key metadata such as titles and publication dates as well as the main text content for each web page were extracted automatically based on structural patterns as well as heuristic criteria on text and markup (Barbaresi, 2021). Important criteria to establish this data source as a basis for further studies were, first, the careful mapping of relevant portions of the web and, second, the ability to extract and preprocess resulting web texts to present them in coherent, clearly described collections. The IT blog corpus is an example of such a data source. For the corpus on German IT blogs, about 700 websites were downloaded, indexed, and organized, yielding approximately 1,507,701 documents and 917 million tokens. The German IT blog corpus is available free of charge upon sign-up at https://www.dwds.de/d/korpora/it_blogs. More information about the structure of the dataset and the blogs that are included is also accessible on this website (Barbaresi and Pohlmann, 2020).

Subcorpus NetzDG (IT blogs)

To analyze the discourse in the IT blog sphere about the German NetzDG, we then compiled a subcorpus containing all blog posts from our overall IT blog corpus that reference keywords connected to the discussion of the law. We searched for the following terms: “NetzDG” or “Netzwerkdurchsetzungsgesetz” or “Gesetz zur Verbesserung der Rechtsdurchsetzung in sozialen Netzwerken” or “Facebook-Gesetz,” or “Lex Facebook.” We assume posts addressing broader issues connected to this law (such as hate speech, overblocking, regulation, etc.) will mention one of the keywords above, as NetzDG has become a symbol of potential overregulation and threats to freedom of expression because of government intervention (Zipursky, 2019). The subcorpus NetzDG allows us to filter and structure the discourse within the IT blog sphere to data specific to this law and helps us identify time frames, bloggers, and additional keywords of potential importance for the discussion. These evaluations can then justify further examination of individual or groups of sources by close reading and statistical analysis.

The subcorpus NetzDG (IT blogs) consists of 738 blog posts that were published in the time between December 2016 and August 2019. Within this time frame, we cover the political discussion culminating in the legislative procedure to pass the bill in June 2017, the law fully coming into effect after a transition phase on 1 January 2018, the publishing of the first mandatory transparency reports by the affected platforms by July 2018, as well as continuing conversations about the effects of NetzDG and its potential revision up until August 2019. The end date of our sample has been determined by the second web crawl for the IT blog corpus. As the corpus will be updated regularly, this end date can be extended.

Figure 1:

Percentage of posts that the respective blogs and websites contributed to the subcorpus NetzDG (IT blogs).

Figure 1 depicts the percentage of blog posts that the respective blogs/websites contributed to the subcorpus NetzDG (IT blogs). It illustrates that a significant amount of blog posts within this subcorpus, about half of them altogether, stem from three IT news portals, in other words, from more professionally organized sites like netzpolitik.org, heise.de, and meedia.de. Smaller outlets (such as golem.de) contributed to the subcorpus too, as did blogs run by interest groups (such as bitkom.org) and individual bloggers (internet-law.de).

Figure 2:

Number of blog posts published that reference the NetzDG per month.

Figure 2 shows the number of blog posts published with reference to NetzDG per month and differentiates each blog by color. We can easily recognize two peaks in June 2017 and January 2018. These points indicate that the coverage about NetzDG mainly followed the legislative procedure since the bill was enacted in June 2017 and came fully into effect in January 2018 after the transition period. Interestingly, the discussion did not stop after the coming into effect of the law, nor after the first transparency reports were published by the affected platforms in July 2018. Instead, the conversation about NetzDG continued at a lower frequency, and we can see that particularly the big three – netzpolitik.org, heise.de, and meedia.de – continued to mention the law at varying levels. Netzpolitik.org engaged in this discussion from the beginning, reporting frequently on the law during the political negotiation process before the bill’s enactment in the months leading up to June 2017. After January 2018, the coverage continued at netzpolitik.org, but it decreased in volume. In contrast, heise.de did not initially pick up the topic but joined the conversation with frequent reporting starting in December 2017. From there on, heise.de continued to publish about NetzDG regularly in the following months.

Newspaper corpus

The second corpus is based on the ePaper editions of the most prominent German daily newspapers made available by the German National Library (DNB). From these sources, we compiled another subcorpus “NetzDG” by searching for our NetzDG keywords. The newspapers in our corpus are: Bild, Der Tagesspiegel, Die Welt, Die Tageszeitung, Frankfurter Allgemeine Zeitung, Frankfurter Rundschau, Handelsblatt, Junge Welt, and Süddeutsche Zeitung. This subcorpus contains every article from these newspapers with one of the above-mentioned keywords concerning NetzDG in the period from December 2016 to August 2019. The corpus allows us to mirror the analyses conducted on the IT blog corpus, and to find out whether and to what extent the discourse on NetzDG differs in these two different media environments.

Traditional print media are still a crucial agenda setter for public and political discussion in Germany. For this reason, newspaper articles are highly relevant to our analysis. Newspapers reach a much broader audience than specialized IT blogs and websites, can prompt discussions surrounding a topic, and provide a platform for the agenda of political decision-makers. In comparison to the very well-informed and profoundly specialized expert discourse happening on IT blogs and websites, the newspaper discourse can represent the discussion in traditional “mainstream” media. Our newspaper corpus contains highly regarded nationwide newspapers with varying political leanings, such as the more conservative Frankfurter Allgemeine Zeitung, the progressive-liberal Süddeutsche Zeitung, or the generally center-left Frankfurter Rundschau. The corpus also includes the tabloid newspaper Bild Zeitung and the conservative Die Welt, both from Axel Springer SE, as well as the left-leaning Die Tageszeitung and the far-left Junge Welt. The business newspaper Handelsblatt and Der Tagesspiegel from Berlin complete the corpus, basically covering the entire landscape of nationwide German newspapers.

Corpus building

To compile the relevant newspaper articles for our project, we drew on the archives of the German National Library, which contain all existing ePaper editions of the above-mentioned newspapers. First, we searched these papers for mentions of the NetzDG keywords within the given time frame. The results of this search were assembled as individual pages in PDF format, taken from the ePaper edition of each newspaper. These are “born digitally,” that is, they were produced without scanning or digitizing the printed newspapers, allowing us to use PDF-to-text tools to extract the textual content without relying on Optical Character Recognition (OCR), a tool provided by the Python library PyMuPDF. The tool’s text extraction feature includes a reliable recognition of text blocks in the given page layout. For each newspaper page, we obtained a JavaScript Object Notation (JSON) file containing the extracted text of each text block along with layout information regarding its X-Y position, bounding box size, fonts, and font sizes. During extraction, PyMuPDF aims to sort text blocks according to their natural reading order on the page.

However, the beginning and end of articles within these pages were neither defined in the PDF data nor easy to detect with any currently available tools. Frameworks which are designed to parse newspaper layouts (layout-parser: https://arxiv.org/pdf/2103.15348.pdf, origami: http://ceur-ws.org/Vol-2723/long20.pdf) offer pretrained neural models for detecting text blocks, headlines, and tables; however, they are not yet constructed for higher-level semantic relationships between text blocks. Therefore, in this project, we opted for a manual solution: annotating bounding boxes on every article on each page using VGG Image Annotator. These annotations were then matched with the positional information of text blocks in the converted data, allowing us to determine which text blocks belong to which article. In addition, we annotated different types of text blocks within an article context for 80 pages: headline, subheadings, image captions, article body text, and author information. Using features based on the layout information given for each text block and a Random Forest Classifier algorithm implemented in the Python library scikit-learn (Pedregosa et al., 2011), we trained a model to predict text block types. This model performs reasonably well, showing F1 scores between 0.79 and 0.84 for individual labels, except for author information which shows an F1 score of 0.49.

Finally, an Extensible Markup Language (XML) file was created for each article containing all the extracted text blocks within its annotated boundaries. Each text block was tagged with its respective type. The order of text blocks was derived from the original output of PyMuPDF. Consecutive text blocks containing article body text were concatenated, including a re-construction of hyphenated words. We verified the quality of extraction during subsequent close-reading of the texts, correcting any apparent errors with comparison to the original PDF files.

Subcorpus NetzDG (Newspapers)

This subcorpus NetzDG (newspapers) consists of 572 newspaper articles, published in the same time frame as the blog posts, between December 2016 and August 2019. The number of articles (572) is relatively close to the number of blog posts (738) in our subcorpus NetzDG (IT blogs) and therefore allows for a comparative exploration of the data.

Figure 3:

Percentage of articles that individual German newspapers contributed to the subcorpus NetzDG (newspapers).

Figure 3 shows that the largest portion of articles on NetzDG within the newspaper corpus, about 32 %, were published by the Frankfurter Allgemeine Zeitung. Five of the newspapers (Süddeutsche Zeitung, Tagesspiegel, Frankfurter Rundschau, Handelsblatt, and Die Tageszeitung) produced around 10 % of the articles each, approximately about 50 articles per newspaper in the given time frame. In contrast, Junge Welt (4.2 %) and Bild (2.8 %) contributed significantly fewer articles to the data set than the other papers, indicating that these newspapers addressed the topic of NetzDG less often.

Figure 4:

Number of newspaper articles that reference NetzDG per month by individual paper.

Figure 4 depicts the occurrence of our NetzDG search terms in published newspaper articles per month and differentiates the individual papers by differing colors. We can identify roughly the same peaks that we have already detected in the blog post graph: Newspaper reporting on NetzDG spikes following the legislative procedure in June 2017 and again in January 2018. As with the findings on blog posts, we can see that the discussion about NetzDG does not completely dissipate after the publication of the first transparency reports (July 2018) in newspaper articles either but continues at a lower level.

Overblocking thesis

We can identify many passages within the discourse that either state that overblocking will happen or is likely to happen due to the implementation of NetzDG. We have called this the overblocking thesis. This argument assumes that due to the short window in which platforms must decide whether content is unlawful (and therefore to be taken down) and because of the high fines associated with the failure to take down content in the expected time frame(s), platforms are likely to delete or block more content than necessary. This argument is essentially an economic one and presumes that platforms will shy away from high fines trouble with government agencies (Hong, 2018). It also implies that platforms will try to minimize costs for content moderation and therefore tend to employ error-prone automatic filtering methods instead of hiring sufficient well-trained personnel for the tasks. At the bottom of this argument lies the idea of new-school speech regulation (Balkin, 2018). It assumes that platforms would be willing to restrict users’ speech (at least rather than risking to get into trouble with government agencies and receiving fines), since it is not their own or their owners’ speech that would get suppressed in such instances. The tremendous amount of data that is constantly uploaded to these sites is often put forward to support the hypothesis that it would be impossible to judge all flagged content in the given time frame(s), especially in cases in which it would be important and necessary to consider the context of the respective speech and contents in question.

Therefore, a typical example of the overblocking argument includes the three following components: It stresses the short time limits for takedown decisions under NetzDG, mentions the high fines that the law imposes if manifestly unlawful content is not taken down on time, and consequently infers, often with reference to the economic interest of the companies, that this combination will lead to an excessive deletion of content.

In view of the vagueness of the proposed law, unrealistically short deadlines (for takedown decisions), and the high fines, this will lead to platform operators being more likely to delete content in case of doubt. The consequence of the draft law would be an orgy of deletion that will also affect a lot of non-illegal content (netzpolitik.org, 2017).

Anti-overblocking thesis

There also exists what we have called an “anti-overblocking thesis.” Supporters of this argumentation also apply economic reasoning; however, in contrast, they propose that it would be in the economic interest of platforms to not delete indiscriminately and immoderately. They believe platforms would try to make nuanced decisions, deleting as little content as possible, because their business model is based on users posting, reading, browsing, and interacting on their sites. Thus, it would be detrimental to their business if they deleted content exceedingly, as they would be risking users becoming frustrated and potentially leaving the platform (Theil, 2018). Such over-removal could also generate a negative public discussion about the company, something these firms surely seek to avoid. A typical example of the anti-overblocking thesis was expressed by the then-sitting Minister of Justice in Germany, Heiko Maas, and taken up in a post by meedia.de:

The operators of social networks have an economic interest in everything that appears on their site. They earn money with every single post, tweet, or contribution. Their economic interest would therefore militate against the assumption that they will now comprehensively delete posts that are not punishable. Social networks also won’t risk losing their users, who would surely turn away if their entries were constantly being wrongfully deleted. (meedia.de, 2017a)

For a better understanding of the discourse about NetzDG in the German public sphere, we wanted to determine, first, the number of occurrences of each argument, second, when they were mentioned, and, third, by whom – in order to see how individual newspapers and blogs positioned themselves with regard to this debate. We also sought to examine whether IT blogs and websites, in comparison to traditional newspapers, as distinct media platforms had differing views on this topic.

Applying machine learning models to quantify the number of appearances of theses in the discourse

We first identified occurrences of each of these theses within a section of the IT blog corpus by close reading in a MAXQDA environment to collect examples of the respective thesis. We then used these examples to train machine learning models to automatically detect incidences of the thesis within the overall IT blog corpus and within the newspaper corpus.

In a first step, all annotated texts were segmented to sentences using the Python library syntok (https://github.com/fnl/syntok). Each sentence was then labeled according to the annotations. If a sentence was not annotated, it was labeled “other.” The Python NLP framework Flair (Akbik et al., 2019) was then employed for fine-tuning the pretrained German BERT model “deepset/gbert-base” (Chan et al., 2020) on the task of classifying sentences. The best performance was obtained after fine-tuning for two epochs, using a learning rate of 1e-05 and a batch size of 16. The results are not outstanding but acceptable given the small amount of data and the complex semantic nature of the arguments. To avoid overlooking many relevant text passages in our subsequent analyses due to false-negative detections, we lowered the threshold of prediction confidence to 0.3 (excluding the “other” class). This resulted in much higher recall and lower precision on the target classes, also leading to many more potential false positives, which we examined and corrected manually. Bearing in mind that the scores for the anti-overblocking thesis are fairly low (F1 score 0.36), we decided to control our results by close reading all blog posts and articles. This verification method allowed us to confirm our results and to analyze the individual occurrences of the anti-overblocking thesis more in-depth in the discussion below.

Figure 5:

Occurrences of the overblocking and anti-overblocking theses in blog and newspaper publications over time.

Figure 5 shows the occurrences of the overblocking and anti-overblocking theses over time. More precisely, it depicts the number of articles and blog posts per month that mention one of the theses. We can clearly see that the overblocking thesis frequently and continuously appears in these publications over time and identify the previously discussed two peaks in June 2017 (enactment of NetzDG) and January 2018 (NetzDG fully becoming effective), which represent heightened interest in this topic due to the legislative procedure. However, we also see that the overblocking thesis is still mentioned long after the first transparency reports were released by the platforms in July 2018. In stark contrast, the anti-overblocking thesis is hardly present in the discourse at all: One can easily spot the infrequent orange occurrences in the graph. The anti-overblocking thesis is only mentioned once in March 2017 and twice in May and June 2017 during the discussion phase before the enactment of the law. Afterwards, it appears once more in October 2017 and then again in May 2019, while the overblocking thesis is present many times in almost every month of the time frame. Overall, we have a total of 294 occurrences of the overblocking thesis in both IT blog posts and newspaper articles compared to just seven occurrences of the anti-overblocking thesis in the same period. Evidently, this is a significant imbalance.

Figure 6:

Appearance of the overblocking and anti-overblocking theses in relation to the number of articles published on NetzDG in a given month.

Figure 6 depicts the appearance of the respective theses in relation to the number of all articles that were published on NetzDG in the given month. We can see that the overblocking thesis is consistently referenced roughly in 20 to 40 % of the blog posts/articles each month. The anti-overblocking thesis, in contrast, is only referenced in at most a couple of articles per month – and often not mentioned in a single article. These two graphs illustrate that the overblocking thesis was very present whenever authors in the IT blog sphere and traditional print media discussed the NetzDG. However, the anti-overblocking thesis is mostly completely absent from the discourse in both media spheres.

Discussion

The graphs and numbers of this analysis provide a quantified overview of the discourse about the NetzDG law regarding the question of overblocking on both IT blogs and websites and in daily newspapers. Notably, the anti-overblocking thesis is almost completely neglected within this conversation, even once the shortcomings of our automatic detection method are considered and eliminated through inspection by close reading. This is surprising, since it is a legitimate hypothesis that has been discussed and argued for by scholars in the field (Eifert, 2018; Hong, 2018; Theil, 2018). Furthermore, there is not much reliable data available regarding take-down decisions by platforms. Therefore, drawing an empirical conclusion as to whether overblocking is occurring (or likely to occur) or not is not possible (Griffin, 2022; Tworek and Leerssen, 2019).^[2] Thus, one would conclude that both theses can only be discussed as hypotheses, since there is not sufficient data available to prove either. Therefore, one would expect that a well-balanced reporting in articles and posts about NetzDG would frequently discuss both of these theses. At the very least, if these articles only discuss one of the two (for example, the overblocking thesis), one expects that the authors acknowledge the existence of the other.^[3] However, our data analysis shows that this is not the case. Instead, we find that the overblocking thesis absolutely dominates within the discourse about NetzDG. It is also remarkable that researchers are not mentioned when the anti-overblocking thesis is discussed on either IT blogs or in newspaper articles. It seems that their voices have not been taken up much in this regard.^[4]

The detected imbalance between the discussion of the two theses is especially noteworthy considering that the anti-overblocking thesis was articulated by the then-sitting German Minister of Justice, Heiko Maas. It is surprising that the views of someone who has such a platform were taken up neither by blogs nor by the traditional print media. These findings raise the question of why the anti-overblocking thesis remained almost completely absent from public discourse in the examined media. What are the reasons for it not being picked up by commentators on blogs or in traditional news media? Why did Minister Heiko Maas and the Department of Justice either not get their message across regarding this argument or why did they deliberately choose to argue for the NetzDG law with different statements? To answer these questions, further research should turn to approaches informed by critical discourse analysis (Fairclough, 1995) and research in strategic communication studies (Hallahan et al., 2007; Thorson, 2013) to identify the underlying rationale for the absence of the anti-overblocking thesis in public discourse as detected in this research. Aside from a detailed examination of the ministry’s communication strategy regarding NetzDG, an analysis of the very effective communication efforts of the Allianz für Meinungsfreiheit, an alliance of numerous civil society organizations, associations, and legal experts against the law (Bitkom e. V., n.d.), promises relevant insights.

A first glance at the publications that mention the anti-overblocking thesis shows that five of the seven mentions are on blogs or websites (com-magazin.de, 2017; meedia.de, 2017a; meedia.de, 2017b; meedia.de, 2017c; netzpolitik.org, 2017). One may assume that IT blogs cover the topic in a slightly more balanced manner. However, given the very low number of posts, this fact does not constitute a real difference in reporting. The anti-overblocking thesis is not well represented on IT blogs either. Turning to traditional German newspapers, only the Frankfurter Allgemeine Zeitung covers the anti-overblocking thesis (one article from 20 May 2017 and one from 4 May 2019). This is equally surprising, as one may expect major nationwide newspapers to inform their readers on a broad range of arguments and perspectives. One may have anticipated that each one of the newspapers would reference the existence of the anti-overblocking thesis at least once over the course of the discourse that we are looking at. However, that has not happened, according to our data.^[5]

Close-reading analysis

At the end of this article, we want to take a closer look at the mentions of the anti-overblocking thesis to identify the conditions under which this outsider opinion found its way into the discourse. The first mention of this thesis in our dataset appears on netzpolitik.org (2017). The blog post from 16 March 2017 gives an overview of the reactions by different stakeholders to the early draft law of the NetzDG. After an initial statement that outlines the website’s own position, representatives from civil society organizations, business associations, platforms (in this case Facebook), politicians from the main German parties, and a lawyer are quoted or their contributions paraphrased. Most of these comments argue against the law or for substantial alterations. However, positive statements are also represented, particularly from politicians of the parties that were in government at the time (the Christian Democratic Union and the Social Democratic Party). An article by the left-leaning Tageszeitung is referenced, which contains a rudimentary version of the anti-overblocking thesis: “Platforms’ current [moderation] practices show that due to their own financial interests, they do not delete controversial pieces of content prematurely” (Rath, 2017). In this sense, the post gives a well-rounded overview of different perspectives on the law. It includes both potential advantages and challenges, including the anti-overblocking argument in favor of the law.

The second mention of the anti-overblocking thesis occurs on 18 May 2017. This was during intense debates about the law before it was going to be enacted by the end of June. The post on meedia.de (2017a) summarizes the main arguments that then Minister of Justice, Heiko Maas, put forward in a Facebook post that day in response to criticism of the law. NetzDG was to be discussed in the German parliament in a first reading the following day (19 May 2017). The post by meedia.de leaves room to discuss the statements of the minister, mostly in direct quotations, among which we find a detailed version of the anti-overblocking thesis. Most significant for our research is the cross-media utilization of content in this case. Maas uses his Facebook account for a justification of his law, which is then picked up and disseminated by an IT blog/website.

The third mention of the anti-overblocking thesis is a Frankfurter Allgemeine Zeitung newspaper article from 20 May 2017, which reflects on the first reading of the law in the German Bundestag. It is generally critical of NetzDG but includes different opinions by quoting or paraphrasing different political stakeholders. Maas’s vindication of the law is covered by paraphrasing a statement that contains a short form of the anti-overblocking thesis: “Maas called concerns for freedom of speech ‘grotesque’. This law would address incitement to murder against unsalaried politicians and journalists, defamation, and Holocaust denial. Facebook would not delete too much because this would cut its own economic interests, said Maas” (Frankfurter Allgemeine Zeitung, 2017). The article is chiefly concerned with the legislative procedure regarding NetzDG and criticizes the short time frame before a decision needs to be made in parliament. It also analyzes the inner workings and interests of the government parties with regard to the law and does not engage in a deeper discussion of the overblocking question.

The fourth mention of the anti-overblocking thesis is on 30 May 2017 in a post by com-magazin.de (2017) that reposts a report by the Deutsche Presse-Agentur. This message provides information on Facebook’s official statement regarding NetzDG (Bundesministerium der Justiz, n.d.) and picks up Maas’s arguments for the law as counterarguments based on his already mentioned Facebook post (18 May 2017). The article contains the anti-overblocking thesis in the form of a fairly comprehensive quote.

The fifth mention of the anti-overblocking thesis appears on meedia.de (2017b) on 27 June 2017, in the final days before the parliamentary decision about NetzDG’s enactment. This is a repost of an article from the Wirtschaftswoche by Marc Etzold (Etzold, 2017) and offers a detailed report of an event organized by the publishing house Gruner + Jahr called “Journalismusdialog.” For this round-table discussion, a representative from Facebook, Patrick Walker, and Minister Heiko Maas were invited to speak with Gruner + Jahr’s CEO Julia Jäkel about “the role of journalism in times of fake news and alternative facts.” Evidently, NetzDG was a main point of the conversation. Arguments from both Facebook’s and the Minister’s side feature in the article, and Maas is paraphrased with a short passage that references the anti-overblocking thesis: “In his [Maas’s] point of view, Facebook has an economic interest in deleting contents either only inertly or even not at all. ‘It’s about pecuniary interests’” (Etzold, 2017).

The last mention of the anti-overblocking thesis from this phase around the enactment of the law occurs on 30 June 2017 in another post on meedia.de (2017c). It consists of an article by the Deutsche Presse-Agentur that summarizes the political and public discussion about the law and briefly reiterates main arguments for and against it. The post is slightly assenting to the law and repeats Maas’s ideas regarding the anti-overblocking thesis in a short paraphrase: “Minister of Justice Maas does not expect any ‘overblocking’ to happen since fines would not be imposed for every violation and since platforms would generally be interested in having more rather than less contents for the purpose of moneymaking.”

Looking at these examples, it is clear that posts from IT blogs and websites contribute to a better balance in arguments for and against NetzDG and regarding the likelihood of overblocking due to its enactment. There is still an overwhelming imbalance of arguments for the overblocking thesis in the IT blog sphere. Nevertheless, we have found few mentions of the anti-thesis. Particularly in comparison to the newspapers (the anti-overblocking thesis was only mentioned once in the Frankfurter Allgemeine Zeitung during the crucial time of the laws’ discussion and enactment), the above-mentioned blog posts reference the anti-overblocking thesis more frequently. However, as argued above, it is debatable whether these few posts could have a big informative impact overall. Therefore, it is fair to claim that the IT blog sphere did not provide much of a platform for a greater diversity of opinions on this particular issue of the NetzDG law. In this regard, the hypothesis presented at the beginning of this article that IT blogs and websites may contribute to a more balanced public discussion about tech-related issues and translate this expert discourse for decision-making in broader society cannot be validated in the specific case of the overblocking discussion. However, further research is necessary to examine whether or not IT blogs and websites were potentially able to provide more and different perspectives on other questions of the NetzDG discourse and were capable of opening up this discussion for non-experts.

With regard to the cross-media development of the discourse and the uptake of certain messages into other media, it is interesting to see that Maas’s Facebook posts represent a platform that allowed him to formulate and position his arguments. Among these arguments was the anti-overblocking thesis, and it is notable that these posts were then at least partly taken up by online media. It is remarkable that only IT blogs and websites reacted to this content, whereas there is no mention of Maas’s posts in newspaper articles. In further research, it would be interesting to see to what extent posts on social media platforms (such as Facebook) by Maas and other stakeholders in the NetzDG discussion have been absorbed, used, and disseminated further by blogs or newspapers. It is noteworthy that neither blogs and websites, nor the newspapers referred to Maas’s Twitter account or any statements that he made there. In this regard, it would be important to analyze Maas’s and other stakeholders’ use of, and discursive practices on, Twitter regarding the NetzDG discourse to be able to draw potential conclusions about the importance, acceptance, and reach of these different platforms regarding the question of overblocking due to NetzDG and the benefits and dangers of the law overall. The “Journalismusdialog” initiated by Gruner + Jahr mentioned earlier, which incorporated the presentation of the anti-overblocking thesis, and the largely positive reporting by meedia.de regarding NetzDG generally indicate the need to do more in-depth studies on the different political and media stakeholders in this conversation and their agendas based on a strategic communication studies approach.

Taking into consideration that NetzDG has been fully in effect in Germany since 2018 (the law was updated in 2021 but not dramatically changed regarding its core elements) and given the situation that only very few instances of overtly aggressive or blatantly wrong blocking by platforms have been noticed by the public (Heldt, 2019), the results discussed above allow for a reconsideration of existing opinions on the matter of overblocking. The relevant scholarship on “over-removal” (Keller, 2017, 2021) and “new-school speech regulation” (Balkin, 2014, 2018) has oftentimes been developed in the US, by analyzing the US media system and drawing on US free speech values (Deibert et al., 2010; Klonick, 2020; Kosseff, 2019). Further research should examine to which extent the NetzDG experience in Germany may or may not represent a (particularly European?) deviation from these important original concepts and findings concerning overblocking, censorship, and platform regulation.

It is by now obvious that Eifert’s (2018) recommendation to take the NetzDG approach as “paradigmatic for the future regulation of social networks” (p. 9) has been accepted and, in fact, been implemented since the European Union’s Digital Services Act (DSA) has taken up core elements of the NetzDG approach (Wilman, 2022). In this sense, the analysis of the initial public discussion about NetzDG (and overblocking) in Germany is relevant for other countries as well since the same general approach is now also applied in other European countries affecting their digital public spheres. It would be worthwhile to examine whether the discourse about the DSA and its NetzDG components was similar in other European countries (and on the level of the EU in general) to the original discussion in Germany. What kind of arguments were used in favor and against the DSA, what kind of stakeholders got involved, and how did they position their arguments and narratives in the debate? Was overblocking still a major issue in this discourse four to five years after the NetzDG had come into effect in Germany, and was it still considered to be a dangerous threat for freedom of speech and liberal democracy, or had the discussion changed by then?