We take up on the idea of a nutrition facts label for online documents: the Information Nutrition Label. Such a label has the potential to increase the readers' ability to make an informed decision before the \consumption" of a news article or some other published online document. The basic ideas along with the dimensions (manifest, measurable text qualities, etc.) of such a label were proposed in [FGG+17]. The paper in hand focuses on the problem of an intuitive, unambiguous, and intelligible label presentation. For this purpose we (1) categorize the originally proposed information nutrition dimensions and (2) interpret them in terms of well-known physical quantities from which we belief that they are intuitively understandable for the general public. To give an impression of our ideas, a visual representation as well as the results of a preliminary crowd-sourcing study are presented.
The World Wide Web is a great source for news. However, relying on online news does not come without di culties both for the individual and for a society as a whole [BMA15]. With the web's sheer endless stream of news on virtually any topic, readers get easily trapped into lter bubbles and may become disconnected from important public discourses. Social networks stimulate the formation of echo chambers where groups of likeminded people share hyperpartisan news while being Copyright © 2018 for the individual papers by the papers' authors. Copying permitted for private and academic purposes. This volume is published and copyrighted by its editors. 17 min 32 °C 31 % 64 dB
B class unaware of alternative arguments and opinions. Furthermore, part of the news articles are spread not only for the purpose of informing people but come with a commercial or a political incentive. Publishers are not loath to use exaggerated or misleading claims and promises in teasers or headlines in order to catch readers' attention (clickbaiting), making it hard to assess the (trust-) worthiness of an article ahead of reading. To improve this situation, the authors of [FGG+17] propose a so-called \information nutrition label" for online news. Like its food counterpart, the label is supposed to help people making more informed decisions upon which news items to consume.
Starting from a set of nine information nutrition dimensions that have been proposed in the original work (see Section 2), we now further shape these ideas towards fewer categories as well as an intuitively understandable representation of the underlying, often complex text analysis results (see Section 3). The proposed label is based on a categorization of the nine information nutrition dimensions into the ve categories I E ort, II Kairos, III Logos, IV Pathos, and V Ethos. The interpretation of these categories shall be simpli ed if associating them with well-known quantities from physics (or nance in the case of V); Figure 1 shows a possible implementation of the idea. I.e., a text is rst analyzed regarding the nine information nutrition dimensions whose resulting values are then combined and rescaled to match the interpretation of the ve categories. In this regard, the value ranges that are employed for the proposed categories were determined with a small crowdsourcing study (see Section 4).
The idea of computing an information nutrition label is an outcome of the Dagstuhl Seminar \User Generated Content in Social Media", held in July 20171, and has recently been published as a SIGIR Forum article [FGG+17]. In their proposal, the authors suggest nine dimensions as attributes for the information nutrition label, with one dimension, \authority / credibility / trust", actually comprising three dimensions into one. Table 1 lists these dimensions in the left column; for a detailed description as well as for a related discussion we refer to the SIGIR forum article. Note that Table 1 introduces two additional information nutrition dimensions, namely \verbosity" and \veri ability", which we see as complementary to the originally proposed ones. While \verbosity" refers to the length of an article, \veri ability" refers to the extent to which an article provides pointers to resources that help to verify claims made [HVER15].
We presume that a label that displays the nine original dimensions (along with their respective statistical measurement units) will receive attention mainly from experts. In order to open the results of an intricate document analysis to the general public (cf. the motivation for a tra c light system to simplify the food nutrition label [foo12]), we ask the question whether a simpler, yet equally informative and hence preferable label can be derived by merging those dimensions that make pragmatically similar statements. 3
Categorizing Information Dimensions When studying the original nutrition dimensions it becomes apparent that some of them, like topicality and virality, or emotion, opinion, and controversy, are similar from a pragmatic point of view. E.g., the topicality 1http://www.dagstuhl.de/17301 and the virality of an article both represent temporal and sociological phenomenons, and both may be used to answer a question like \How much do others care about the article?". Because of correlations like these, we presume that the nutrition dimensions can be subsumed into categories without a signi cant pragmatic loss of information.
In the second column of Table 1, our proposal for such a categorization of the altogether 13 dimensions into ve groups is shown. The labels for the categories II . . . V have been chosen in accordance to Aristotle's modes of persuasion [AK07], including the less well known concept of \kairos", which stands for the \right, critical, or opportune moment".2 Column 3 shows our proposal for a category interpretation in physical terms ( nancial in case of V), along with sensible value ranges in Column 4. In Column 5, a prototypical user question which is addressed by the respective category is stated. In the following, each category is discussed in detail.
Category I, e ort, groups all dimensions that a ect the time a reader has to allot to comprehend an article. Besides verbosity, which has been used already by media websites to provide an estimated article reading time (e.g. by Medium3), the dimensions readability and technicality fall into this category. To express e ort, we consider time in minutes as an intuitive choice. The e ort category allows readers to check whether they have enough time to read an article and to identify articles of a speci c depth.
Category II, kairos, groups all dimensions that pertain to the trendiness, momentum, or hotness of an article or a topic, i.e., topicality and virality. As an existing attempt to provide this category, the velocity graph [Pet13] on the media website Mashable4 can be 2https://en.wikipedia.org/wiki/Kairos\#In_classical_rhetoric 3https://medium.com 4https://mashable.com counted. As a quantity to express kairos, we consider the temperature in the range of 0 100°C as intuitive. The kairos category can bring articles to readers' attentions which would be \out of their bubble" otherwise.
Category III, logos, groups all dimensions that capture how well an author supports her claims with evidence, i.e., factuality and veri ability. As a quantity to express logos, we consider transparency in the range from 0 - 100% as intuitive. The logos category can help readers to assess the journalistic quality of an article up front.
Category IV, pathos, groups all dimensions that are related to subjectivity and discrepancies, i.e., emotion, opinion, and controversy. As a quantity to express pathos, we consider volume, measured as sound pressure, as intuitive. The pathos category can help readers creating awareness that communities sharing alternative arguments or opinions likely exist.
Category V, ethos, nally groups all dimensions related to the credibility of an author or publisher, i.e., authority, credibility, and trust. As a quantity to express ethos, we consider credit ratings in the range from A+ to D, as used in nance,5 as an adequate choice. The ethos category can help readers assessing the risk of becoming misinformed or, alternatively, the potential of learning about non-mainstream viewpoints. 4
We see three advantages when using the proposed categories as attributes for the envisaged information nutrition label instead of the original dimensions. First, the reduced number of attributes makes the label both easier to present and easier to digest in practical settings. Second, by resorting to well-known quantities for the categories, readers can intuitively interpret the label without the need of detailed instructions. Third, the chosen quantities allow for the design of a non-textual visualization of the nutrition label. On the other hand, the potential concerns should not be overlooked: rst, the categorization may be not as lossless as anticipated, such that the ve categories convey much less helpful information than do the original dimensions. Second, the quantities (or their visualizations) may lead to false intuitions about the document they belong to.
As a very rst step towards clarifying some of these concerns, we have designed a mainly non-textual representation for our label (see Figure 1), which allows for a visual comparison with the tabular label presented in the SIGIR forum article. In Figure 1, each category is visualized by a rounded rectangle featuring a category symbol and an article-speci c category value. For the latter, both the absolute value as well as its relative 5https://en.wikipedia.org/wiki/Bond_credit_rating\#Credit_ rating_tiers 0.5 0.0
Effort
Kairos
Logos
Pathos
Ethos position in the value range are depicted. To make the values in the depicted label more sensible, we asked a crowd of 42 workers to read and then annotate the news article exempli ed in the SIGIR forum article in the light of our categories.6 As nal value for the label we took the mean of all 42 annotations. For the sake of completeness, a violin plot showing the distribution of all annotations is shown in Figure 2. The violin plot indicates that the above article takes little e ort to comprehend and, it is obviously not very hot anymore (kairos). In terms of transparency and sound pressure (logos and pathos), no clear consensus is reached, while the publisher is clearly not top rated in terms of credibility (ethos).
For future work, we consider to present a
computational model for the information nutrition label and
to further investigate the correlation of its constituent
parts with human intuition.
[AK07]
[foo12]
[Pet13]
foodwatch. Research supports tra c light colours.
https://www.foodwatch.org/more-information/
research-supports-traffic-light-colours/,
2012.
Rockmore. Measuring veri ability in online information. CoRR, abs/1509.05631, 2015.