Algorithms in digital media and their influence on opinion formation
Background and central aspects of the topic
Online media – which have been used by more and more people in recent years to get informed on important societal topics – are a major source of political information and serve as a basis for opinion formation. In this context, just like Internet portals and social networks, search engines nowadays are an integral part of the everyday media use of many people.
The functioning of digital media is characterised by algorithms. Thus, for example, algorithms decide which websites are considered to be relevant for a search query and in which order the results will be presented. Based on data concerning media use, they determine which selection of messages (and ads) will be displayed for a user and when it will be displayed. This is how algorithms can have an influence on opinion formation.
However, not only social media, but also conventional media coverage is increasingly characterised by algorithms. Journalists – just like further stakeholders who contribute to opinion formation – are supported by algorithms when it comes to key tasks such as research, weighting and selection of information or the production and distribution of articles etc. Machine-generated messages are still limited to areas for which standardised data (e. g. sports tables, company reports) are available. In view of the rapid development of machine learning algorithms, however, an extension to other areas is to be expected.
The use of algorithms also is intended to explore preferences of the recipients: Thus, for example, the Washington Post or the US news portal Upworthy show test persons different versions of their articles prior to publication. A software then determines which combination of titles, images and text modules are accepted best by the readers.
It has to be assumed that algorithms will play an increasingly important role for opinion formation. On the one hand, their use enables media consumers to access a wide range of information and (even political) statements more easily compared to pre-digital media. On the other hand, there is a risk of opinion formation being influenced or even manipulated by algorithmic mechanisms that very often are not clearly identifiable (e. g. the Facebook news feed algorithm or search engine ranking).
In recent years, the interaction between algorithms and digital media has moved into the focus of attention – first among experts, but increasingly in the mass media as well. In this context, i. a. the following questions are being discussed: Which content is displayed for people using Facebook and which lists of results are displayed in a Google search? How do media providers make use of the new opportunities e. g. for prioritising and creating media reports? What is the significance of algorithmically supported media services for opinion formation or generating (political) awareness? Do fake news, hate comments or »filter bubbles« have an impact on democratic decisions?
Nevertheless, in 2016, an interdisciplinary team of scientists outlined in the scientific journal »Nature« that there are still too few analyses regarding the (possible) impacts of using algorithms on social, cultural and political areas. Civil society initiatives call for at least a better transparency and (if necessary) control of algorithms.
Objectives and approach
The TAB project shall approach the complex topic from two directions: The dynamic developments characterised by algorithms shall be discussed with regard to both conventional and social media. Influences on opinion formation shall be focused on. Literature analyses with regard to the technical basics and the use of algorithms in digital media, scientific discussion and potential political instruments shall serve as a starting point. The public debate on these issues shall be reflected by an analysis of contributions in the press and in broadcasting. Interviews with experts from science and practice shall complement the approach.
Selected aspects shall be dealt with in depth within the framework of dialogue elements of the Stakeholder Panel TA. For this, discussions shall be held in focus groups and theses shall be generated, e. g. with regard to the questions of which potentials, limitations, risks and opportunities users associate with personalised news and messages. The results from the focus groups shall be validated in an online survey.
After an extensive literature and media analysis, expert interviews were conducted and an expert opinion was commissioned on the topic of "Effects on individual opinion formation among users," in which personalization algorithms are analyzed in depth and their significance as part of information intermediaries is clarified. The report summarizes the state of research on the effects of algorithmic processes in opinion-forming processes and discusses implications for media law.
In order to collect the views of societal stakeholders, an online survey was conducted in the spring of 2018, in which the evaluation of increasingly automatically personalized news offerings, the advantages and risks of such offerings with regard to opinion formation, and suggestions for strengthening the diversity of opinions and topics were asked. In addition, the views and attitudes of young people toward personalized online media were surveyed in advance with individual and group interviews and a representative nationwide survey. The results were published in TAB Sensor No. 1 ("How do young people evaluate personalized online media?", see below).
Taking into account the continuously emerging scientific publications in the field, the various results of our investigation were combined in the final report, which is currently going through the process of approval by the group's TA rapporteur group.
Kluge, J.; Oertel, B.; Evers-Wölk, M.
2018, August. Büro für Technikfolgen-Abschätzung beim Deutschen Bundestag (TAB)