Conceptualizing different forms of news processing following incidental news contact: A triple-path model

Processing assumptions in path A: Automated usage leads to dominance of fast thinking

Sundar (2012) argues that users have more control over their news consumption in digital media environments which is supposed to increase involvement and thus the motivation to learn. How does this correspond to news consumption through newsfeeds? In principal, users here, too, have the opportunity to actively individualize their news consumption by means of explicit personalization (Bozdag, 2013), for example, by following certain profiles. However, this act of active control is probably only dominant when users adopt a new platform.

LaRose and Eastin (2004) argue that users may make a conscious decision that corresponds to their expected gratifications when using a new medium initially, but over time develop habits that no longer directly respond to their initial needs. Hence, in the early phase, users repeatedly adjust their newsfeeds until routine use reveals satisfactory results. Continuous use of the platform might result in overlearning the initially active in favor of a more passive and relaxed approach. Moreover, though personal curation of newsfeeds is until now only discussed in the light of selective exposure (e.g. Thorson and Wells, 2016), it is also reasonable to assume that self-selection of content likewise influences the processing itself: By actively adjusting their newsfeeds, users in the long run create streams of ‘likeable’ news posts, making newsfeed usage a highly pleasurable experience; ‘just scrolling’ feels rewarding and becomes routine. With this automation in usage, less mental resources are needed for decoding, but also less mental resources are made available by the user because the nature of the newsfeed content does not suggest any need for sophisticated processing (Kruikemeier et al., 2018; Salomon and Leigh, 1984).

This corresponds well to the characterization of habit by repetition, but habit is furthermore defined by ‘deficient self-regulation’ (LaRose and Eastin, 2004). It, therefore, seems plausible that especially routine newsfeed users fall into automatic mode, in which conscious self-regulation is not active, at least at first:

A1: When using newsfeeds routinely, overlearning occurs, which means that behavior is more and more automated. As a result, less mental resources are needed, but also less are attributed.

Dual System Theories (DST) of human information processing are widespread. By contrast, Buijzen et al. (2010) propose a third route to describe possible effects of persuasive messages: While in many DST, the processing in intuitive mode is primarily associated with low levels of attention, in this third automatic mode, users are also completely unconscious of the fact that persuasive messages are actually present.

Furthermore, an automated newsfeed usage out of habit implies a dominance of system 1, ‘manifested behaviorally in cognitive-emotional preoccupation with using SNS’ (Turel and Qahri-Saremi, 2018: 3053). Groot Kormelink and Costera Meijer (2019) indeed find evidence for such an impulsive drive when examining user perceptions on scrolling: ‘What the logic of “scrolling” (i.e. “keeping it moving” even if one likes the content) suggests is that users find it (increasingly?) challenging to “commit” to one news item when there is a plethora of other content to consume’ (p. 12):

A2: The newsfeed’s affordance for continuous scrolling makes the top-down allocation of attention more difficult, leading to a dominance of the intuitive mode of thinking. (system 1)

News posts within newsfeeds usually comprise different modalities, that is textual and (audio-)visual information. Building upon the notion of verbal–visual redundancy, it is often assumed that information presented in multiple modalities is more likely to ‘getting through to receivers’ (Sundar, 2000: 482). This, however, would presume that users are indeed engaging with the textual information, which is, as we argue, not inevitably given for news consumption through newsfeeds. Moreover, recent findings from psychology and neuroscience do point to a predominance of visual over textual information regarding the ease of processing (Powell et al., 2019). Especially for news posts, eye-tracking reveals that it is richer content (i.e. pictures and hyperlinks) that enhances attention (Kruikemeier et al., 2018).

This also refers back to the ‘cognitive-emotional preoccupation’ (Turel and Qahri-Saremi, 2018) when using social media. Since the unique quality of visuals is to be eye-catching and easily perceivable, they ‘exert an activating effect by fostering an emotional connection with the reader’ (Powell et al., 2019: 573). Given the hierarchical nature of news processing from attention to elaboration as stated by the CMM, it is more likely the visual aspect of a given news post that guides the preconscious filtering of content in the automatic mode. This in turn might lead to a greater influence of emotional responses when news posts within multimodal newsfeeds are selected by users in automatic mode:

A3: A dominance of the intuitive mode of thinking when scrolling through the newsfeed in the automatic mode manifests in a greater influence of emotional responses when selecting news posts for closer attention (on path B or C).

Learning potential of path A

In the field of consumer research, studies of incidental (ad) exposure as well as of implicit effects are widespread, as research on advertising effects to a large extent cannot assume active processing. In political communication science, implicit effects are less investigated, although they can explain, for example, the effectiveness of political campaign banners on the basis of priming, that is, by temporarily increasing the accessibility of a mental construct (Buchanan, 2015: 323). This can also be done ‘subliminally’, without attention.

Against the background of the typical use of news outlined earlier (teaser level, scrolling by, without an explicit information goal), the contact conditions primarily applied to advertising stimuli in the past now also appear realistic for omnipresent news. Indirect effects of news would be of interest as news-related knowledge structures can be unconsciously activated. This, in turn, can stimulate appropriate behavior (e.g. to recognize and heuristically process a news post when encountering one during further newsfeed usage, see path B). Although implicit processes cannot lead to the formation of new goals (see Knoll et al., 2018, on implicit goal activation), implicit knowledge can thus promote the acquisition of factual knowledge.

The accumulation of different cues (e.g. technological interactivity cues, Sundar, 2012, or emotional cues) while scrolling enables a ‘ready application of cognitive heuristics triggered by those cues’, leading to information processing likely biased ‘toward heuristic, rather than systematic, processing’ (Kim and Sundar, 2016: 48). Hence, the longer people are scrolling through the newsfeed, the more cues are activated, making heuristic processing more likely than a systematic examination of a specific news post on path B:

A4: The automatic use of the newsfeed activates cues without this being conscious to the user (unconscious priming). As a result, heuristic processing of the content (path B) may take place if the cue is encountered again.

On a purely physical level, it can sometimes be difficult to make a clear distinction between gradually slower scrolling and mental dwelling on a particular news item. Nevertheless, we assume that this differs in terms of processing: Moving beyond the rather rudimentary classification of system-1 versus system-2, we follow Stanovich (2009) in proposing that very different type-1 processes do exist and that path A and path B are variations of the ‘Autonomous Set of Systems’ (p. 56). This enables the triple-path model to consider that both the scrolling through the newsfeed as well as stopping for brief glances at a news post are not active forms of processing, but still differ: Heuristic processing in path B is more associated with explicit memory, and though not made consciously, the result can still be verbalized (Stoker et al., 2016). By contrast, scrolling without focused attention (path A) is more an implicit processing, contributing to implicit memory and not leading to verbalizable outcomes.

Here, we urgently need empirical research on automatic processing on social media to better assess the effects of the emotional, subconscious reactions particularly to visuals in newsfeeds. For example, Müller et al. (2016) show that the number of perceived political Facebook posts leads to a greater feeling of subjective knowledge. The authors explain this with the activation of stored knowledge structures, implying that news use has taken place. Since actual knowledge is not acquired to the same extent, these ‘false heuristic inferences’ go hand in hand with increased ‘illusions of knowledge’ (Müller et al., 2016: 439). It is currently an open question whether users form the perception of having seen ‘political’ posts based on automatic processing (e.g. by scrolling through pictures of Donald Trump) or whether this requires the heuristic processing of path B. Moreover, given the important role that emotions play in processes of selective approach and avoidance of news (Song, 2017), it may be that the emotional cues accumulating on the automatic path already contribute on a subconscious level toward directing attention to partisan news items which are then processed heuristically or selected for systematic perusal on path C:

A5: Given the repetitive nature of news encounters within newsfeeds, the affective imprints of a news posts, largely based on visual or single word cues, may already contribute to ‘illusions of knowledge’ as well as selective approach and avoidance.

Processing assumptions path B: Heuristic evaluation on teaser level

These attentional responses can lead to visible social interactions, often operationalized as active engagement (e.g. Karnowski et al., 2017; Oeldorf-Hirsch, 2017), or might be restricted to ‘invisible’ mind focusing. Although news content at the teaser level is limited in terms of substance and complexity (Kümpel, 2019), these ‘snack news’ can in principal be processed with various levels of effort. However, ‘news snacking’ is not an attentive form of news reception: Though the news post grasped the user’s attention at first glance, it has not exceeded the personal relevance threshold which would have led to actually clicking on the linked article and thus to attentive, systematic processing in path C. Instead, we assume that processing based on heuristics is more likely.

Put simply, heuristic information processing is based on learned knowledge structures derived from experiences, observations, or socialization (Chaiken, 1980). These knowledge structures provide processing rules for specific situations, enabling an efficient and resource-saving form of processing. Heuristic processing is used particularly when only a limited amount of information is available or when the motivation is not sufficient to systematically process a situation and its content. Accordingly, heuristics are rather simple explanatory models.

The prerequisite for applying the learned and stored heuristics, however, is that they are present in memory, in other words cognitively available and activated. Activation happens by cues (e.g. while scrolling); if a certain cue is recognized, the corresponding processing rule is activated. An example for heuristic processing is the ‘proportion of news in feed’ acting as a cue for the importance of a topic (Stoycheff et al., 2018: 182). For this ‘peripheral agenda-setting’, media (or technology) cues are found to be more relevant than the content itself (Bulkow et al., 2013).

In most cases, people are not aware of how they have formed their opinion or knowledge: Content can be processed on the basis of learned rules (heuristics) so that the actual information is not stored beyond working memory, only the ‘evaluative implications’ of that information are recorded (Kim and Vishak, 2008). Instead of the message itself, an affective imprint of message exposure remains (Graber, 1988).

Learning potentials of path B

Even in traditional contexts, headlines not only need to convey a summary of the article, but also to attract attention to encourage reading (Andrew, 2007). The newsfeeds’ structure as an endless stream of potentially interesting, personalized content reinforces this competition for the scarce good of attention. Hence, headlines tend to be formulated in an extremely pointed or deliberately ambiguous way, often emotionalizing through negativity. Headlines can thus be misleading without being directly false (Ecker et al., 2014). When snacking news, these headlines serve as cognitive shortcuts for users, enabling them to efficiently deal with the rich supply. However, those who stay at the teaser level are provided with a different set of heuristic cues compared to those reading the full article (Andrew, 2007).

Findings from Yadamsuren and Erdelez’s (2016) qualitative study underline that especially the unusual, the bizarre, the outrageous or the crazy lead to allocation of attention after unintended contact with news, implying that soft news are more likely to succeed. As users with low interest in news are already less likely to find hard news in their newsfeed due to the different curation modalities (Kaiser et al., 2018; Thorson and Wells, 2016), this would further strengthen the dominance of soft news in INE. This is supported by findings for traditional media, according to which visual cues and other salient features of a news article determine the allocation of attention resources for further processing of the content (Bucher and Schumacher, 2006):

B1: Given the automated relevance checks that occur during scrolling, news posts resembling soft news (strong visuals and emotional cues) are more likely to capture attention within the competitive newsfeed. Hard news (if present at all) are less likely to be recognized as relevant.

Furthermore, news teasers in newsfeeds are dominated by visuals and videos. As evidence from framing research shows, in the presence of both visual and textual information, the visual frames have a stronger impact during heuristic processing (Powell et al., 2019), even though the difference may depend on individual characteristics such as the processing style:

B2: Given the dominance of visual over textual information during heuristic processing, knowledge gains based on visual frames are more likely on path B.

Lee and Kim (2017) find that recognition of political information is, unlike recall, not dependent on actual exposure to the content. That is, even when examining a news post superficially, it may still contribute toward recognizing particular aspects. However, this recognition potential is restricted to the teaser level in path B. Hence, reasonable learning outcomes based on path B are the recognition of political actors or other forms of rather simple factual knowledge, but not the recall of complex factual knowledge (Kümpel, 2019):

B3: If a news post gained attention, it is initially processed based on heuristics, leading a) to positive effects on recognition but not on recall. Since exposure is restricted to the teaser level, recognition is b) limited to simple facts.

However, the formal characteristics of a news post also act as peripheral cues that drive social interactions in form of liking, commenting, or sharing (Oeldorf-Hirsch and DeVoss, 2019). Especially the liking of news posts is a low-effort and low-threshold form of engaging with news content (Knoll et al., 2018), which still implies concentrating attentional resources for a certain amount of time. It is thus plausible that ‘the more individuals elaborate on a snack news post, the more likely they may become to click on the link and seek more in-depth information [. . .]’ (Heiss and Matthes, 2019: 4). Moreover, active engagement is acting as implicit feedback for algorithmic personalization (Bozdag, 2013) or might encourage friends to share more of these news posts. Heuristic interactions with a news post might thus increase the amount of news posts within the individual newsfeed. We thus assume that active engagement on the teaser level might lead to a ‘gateway-effect’ (Baum, 2003), that is reaching inattentive audiences by making information more accessible:

B4: Engagement with a news post indirectly increases the proportion of substantial news posts by means of algorithmic or social curation.

Processing assumptions in path C: Goal activation leads to systematic processing

As suggested by Knoll et al. (2018), we assume that goals do not always have to be pursued in the foreground and are not always conscious (Dijksterhuis, 2013). Rather, they guide our behavior because they direct our attention to goal-oriented information. After encountering a news post on the teaser level without initial information needs, the cursory examination of the teaser might reveal that it is related to an underlying goal and thus lead to it being called up.

According to uses and gratifications research, surveillance needs and social utility are the most important news use motivations in terms of knowledge gains (Eveland et al., 2003; McLeod and Becker, 1974). Hence, calling up news-related posts occurs predominantly when there is a thematic interest (Karnowski et al., 2017). However, such an ‘interest’ in a given news item might also be generated in situ (Kümpel, 2019) through perceived informational utility or emotional and thus more impulsive reactions. Another important determining factor for news engagement is the news recommender: social curation might increase the social relevance that is attributed to a given news post (Kaiser et al., 2018). Addressing the news user in person, through tagging or direct speech, increases the probability of active news engagement (Kümpel, 2019). Thematic prior knowledge, however, impedes reading (Karnowski et al., 2017):

C1: Calling up the full article depends on the activation of an underlying interest in the news item and thus of a perceived informational, social, or hedonistic relevance.

If the news item finally exceeds the personal relevance threshold, the resulting active news engagement corresponds with ‘common’, active learning assumptions. The more a recipient is interested in a news item for informational needs, the more she is attributing attentional resources to it and the more likely she is to elaborate on the content. Given that INE is initially not associated with news interest, this path C is the path with the most prerequisites and the most unlikely to occur. This is confirmed by the empirical observation that an average user only clicks 7 percent of the political news stories available in the personal newsfeed (Bakshy et al., 2015).

Learning potential of path C

Lee and Kim (2017) demonstrate that active learning through news engagement after unintended news contact can lead to a positive and moderate effect (B = .51***) for the acquisition of political knowledge, which is remarkable for media effects. This effect is similar to that expected for reading newspapers, which is plausible because ‘the act of reading nearly forces attention’ (Eveland et al., 2002: 363):

C2: Due to the necessary condition of perceived relevance, calling up the article after incidental news contact can be treated as intentional news use, leading to active processing and thus to factual knowledge gains (according to the substance of the actual news content).