Information scent – credibility and gaze interactions: an eye-tracking analysis in information behaviour

Introduction

The online environment or the Web offers a limitless plethora of information for the fulfilment of search goal of information seekers.  From the reputable institutional sites to the narrative personal blogs, users are privileged to draw on the varied visual cues and links to guide their navigation. These elements are generally described in various studies as source cues, visual cues or informational cues. They are conceived as ‘prey’ in information foraging concept (Pirolli and Card, 1999), and they emit an information scent to assist users to move from one webpage to another efficiently.  Individuals may rely on the information scent in applying evaluative judgement (see Rieh, 2002) or preferences to assess the value, quality and credibility of the information patches (websites or webpages).

Past research on credibility relies on the descriptive identification of linking this perception to the source (Eagly and Chaiken, 1993; Rieh and Belkin, 2000), authority (Fogg, 2003) and value such as trustworthiness (Olston and Chi, 2003) and users’ predictive assessment (Rieh, 2002). It is important to examine the behavioural aspect of information seeking in how users expend their effort and energy with their concept of credibility in different information patches. As information scent presents itself in different forms, for example in strong-scent, weak-scent or no-scent patches (Sundar, Knobloch-Westerwick, and Hastall, 2007, p. 1), information seekers may adopt different techniques in their responses to the visual elements displayed in the online environment. Humans are adaptive in their information searching behaviour and often adopt flexible attentional system (Tomasello, 1999; Blair, Watson, Walshe, and Maj, 2009) to adjust their allocation of focal attention relevant to their task goals and cognitive information processing.

Hence, the goal of this research is to extend the concept of credibility and information scent in how users express their information behaviour through gaze interactions while foraging for information.

The study employed an experimental method to investigate users’ information search behaviour and attentional focus relative to the different information patches (webpages). An eye-tracking mechanism was used to examine participants’ eye movements and gaze interactions (scanpath). Post-interviews and surveys were conducted, using quantitative and qualitative analysis.

Literature Review

The concept of information scent and cues

The ubiquitous web environment inundated with repository of information requires the discerning online users to make selective information decisions.  Information seekers often rely on the online hypertextual links or information scent links to help them navigate the web environment

Information scent and cues play a pivotal role in connecting information sources and users through information spaces. Information behaviour researchers had referred to the online environmental cues as visual cues (Tomasi, 2014), proximal cues (Pirolli,1997; Pirolli, 2003), source cues (Winter and Kramer, 2012), and informational cues (Lee and Pang, 2014). Other intrinsic cues such as source expertise, guides the users’ specific search on relevant topic (Petty, Cacioppo and Goldman, 1981). Expertise cue relates to the profession and status of the author of a blogpost influences the perception of credibility of the source. Burgoon and Saine (1978) identified chronemic or temporal cues as non-verbal cue system that shapes the perception of individuals. 

Information foraging theory posited that cue transmits an information scent by which people use as the proximal means to judge the value and relevance of distal information sources (Pirolli and Card, 1999). These cues influence users’ heuristic approach in assessing elements that are relevant to their information needs (Sundar et al., 2007).

According to the concept of information foraging, cues left a trail, be it overt or covert, and emit as information scent that guide information searching and processing. Pirolli and Card (1999) drew parallel analogues to the way organisms hunt for food or prey in various patches is akin to human’s information retrieval behaviour. The salient characteristics rest on the users’ strategies and techniques to maximize the cost and rewards, with the various alternatives and options available within the patchy environments (Stephens and Krebs, 1986). Central to the theory of information foraging is the perceived appeal of the information patch (Pirolli and Card, 1999; Blackmon, 2012).  Card, Pirolli and Van Der Wege (2001, p. 2) suggested that information scent may be ‘delivered by link descriptors, images, contextual clues, for example headings or page layout’. When users forage for information, they may traverse between patches, for example by clicking on an image or a textual link to view further information or move to other patch, if the current site is scarce with relevant content.

The extent of information scent derived from cues result in users perceiving the information patch with ‘strong scent, weak scent, or no scent’ (Sundar, Knobloch-Westerwick, and Hastall, 2007:1), as the scent shapes users’ judgement and processing of information.  The information scent may be interpreted as prey for users (Pirolli, 2007) during information processing, and where users purposefully select cues that offer most value (Kalyanaraman and Ivory, 2009). 

A website with strong informational cues determine users’ attribution of the popularity and credibility of the source, objectivity of reviewers (Sundar, 1999), text and images, captions and hyperlinks (Yaari, Baruchson-Arbib and Bar-Lian, 2011). Sandstrom (1994) highlighted that these paths optimally assist online users in their search to envisage easy retrieval of information. 

Previous research on source cues (Winter and Kramer, 2012) postulated that the user-generated contents and the expertise of the author of a blog post determines their source credibility (Wilson and Sherrell, 1993). This judgemental value of information, amid covertly, possess a strong scent. Other studies (Schamber, Eisnberg and Nilan, 1990; Tombros, Ruthven and Jose, 2005) suggested that user’s appraisal of information quality and relevance is significantly associated with cues such as therecency of information and authorship. 

On the contrary, weak informational cues such as lack of popularity, absence of reviewers’ comments, poor use of language, unprofessional pixelated images may lead users to accord low level of credibility of the information patch.  In optimising one’s decision during information foraging, individual may decide to leave the less sticky, and low information scent patch.  McCart, Padmanabhan and Berndt (2013) extended the characterisation of information scent to strict and relaxed information scents. This phenomenon conveys users’ heuristic information behaviour instead of systematic approach of processing information (Chaiken, 1987; Sundar, Knoblock-Westerwick and Hastall, 2007). 

Studies on information foraging theory (Schamber et al, 1990; Sundar, Knoblock-Westerwick and Hastall, 2007) and information scent have investigated users’ behaviour of switching patches. There was scant research on the users’ attentional focus relative to the varied information scents, and how users engage these scents in information patch of varied genre.

Information credibility

The salient characteristics of credibility defined by Hovland, Janies and Kelly (1953) are trustworthiness and expertise. Other scholars have extended the definition to include reliability, fairness, objectivity (Self, 1996), persuasive sources, message structure and content, media perception (Metzger et al., 2003; Flanagin and Metzger, 2007).

Informational seeking activities entail the engagement of individual’s cognitive thought process through assessment, rendering and interpretation of information (Pirolli and Card, 1995, 1999). In the context of optimising the information search goal, individuals commonly adopt a satisficing decision, using rules of thumb instead of a rational choice (Gigerenzer and Goldstein, 1996) during information processing and retrieval. The process of discerning the perceived value and credibility of the informational cues or expertise of the source requires a series of evaluative judgements on the part of the users. Users’ augment their information search through assessing the value and quality of the surrounding cues (Olston and Chi, 2003). Information seekers’ information selection is assessed based on the credibility of the user-generated contents of online information (Winter and Kramer, 2012). For instance, the source’s website sponsors and website design may project a strong scent or weak scent relating to perceived credibility of the site (Flanagin and Metzger, 2007; Westerwick, 2013). 

Past research suggested that persuasive influence of a credible source is more significant than a non-credible source (Eagly and Chaiken, 1993; Petty and Cacioppo, 1988). In information decision making, the popularity of the source and institution (Rieh and Belkin, 2000), familiarity of the topic, language of the writer, layout or colour of the websites, influence the user’s ‘credibility judgements’ (Gunther, 1988; O’Keefe, 1990; Westerwick, 2013). Fogg (2003) suggested three main source of web credibility includes operator (i.e. organisation or author of the site), content and design. Fogg postulated that the ways users assess credibility are by noticing due to prominence, and follow by their interpretation at the starting point.  Rieh (2002) adopted an iterative approach in the credibility assessment process and suggested that users expressed preferences in tapping on predictive judgement to evaluate credibility evaluation. Using ground theory method, Hilligoss and Rieh (2008) further examined credibility at three levels - construct, heuristics and interaction.  

Source credibility denotes predictive judgment (Rieh and Belkin, 2000) in which information seekers engage in during information processing, prior to visiting the web sites, while evaluative judgement occurs during interactions with the webpages. Choi and Stvilia (2015) pointed out that topic familiarity (Arazy and Kopak, 2011) and users’ information literary levels (Ahmad, Wang and Hercegfi, 2010) play a significant in their credibility assessment of websites.

Extending beyond the source credibility cue, this paper aims to analyse attentional focus and information scent during evaluative stage of information processing. When retrieving information, the emergence of informational cues and links residence within the websites may contribute to the users’ judgemental evaluation of contents or messages created by the source.

To our knowledge, studies were scarce on linking users’ evaluation of information credibility, the condition of task relevant to the engagement of information scent, and the effort and energy (e.g. time) incurred on different information patches. This is significant as in the process of foraging for information, users tend to weigh the cost and benefits and apply preferences and judgemental attitudes in engaging proximal cues to fulfil their information task goal. This research contributes to better understand the concept of information scent theorised in the information foraging theory.

Based on this framework, this leads us to the following hypotheses and research questions:

• H1a: Users will evaluate information patch with cues that emit strong information scent a high level of information credibility. 
• H1b: Users will evaluate information patch with cues that emit weak information scent a low level of information credibility.
• RQ1: Do types of task influence the users’ perception of credibility?
• H2a: Users’ perception of website with strong information scent lead them to spending a longer duration of time (attentional focus) on the information patch.
• H2b: Users’ perception of website with weak-information scent lead them to spending a shorter duration of time on the information patch. 

We reason that if the source or institution of the website is reputable due to their brand popularity or expertise status, including professional articulation of content, these elements will project themselves as positive information cues, and thus users will perceive the information patch as credible. This research includes analysis on whether users will spend a longer time on websites with strong information scent or weak information scent sites. The assumption is based on the premise that on websites with strong information scents display more appealing and useful contents, and exhibit appropriate hyperlinks that assist information search on the web, than websites with weak information scent. Thus users will allocate a longer time reviewing contents derived from the informational cues. On websites that possess weak information scent, individuals may experience difficulty in retrieving and interpreting information.  Users will tend to scan through the pages and dispense shorter time in viewing the site due to the absence of useful resources and cues, and resulting in them exiting the site faster.

Task and attentional focus

Blackmon (2012)’s study on task difficulty and user’s allocation of attention within the webpage were generated from the links-clicked data. Their findings reflected that participants who took on difficult tasks were less successful in task completion. Their experiment showed that a higher number of links, for example 93 links on webpage leads to a higher success rate of task completion. However, Miller and Remington (2004) posited that eight links per webpage is ideal. 

Liu and Belkin (2014)’s study on task and users’ writing and searching behaviour, found that users with higher topical knowledge showed no difference in the two tasks, and users with lower topical knowledge spent longer time searching.  These studies implied that task context such as structure of the task and links mechanism may influence users’ information searching behaviour.

Attentional focus using eye-tracking analysis

Information seekers use the filtering approach that enables them to maximise their information utility.  According to Bates (1989), ‘berrypicking’ is another heuristic approach throughout the search process (Ingwerson and Jarvelin, 2005).  In encountering task or problem situation and the availability of the external information attributed by information seekers, they may potentially adjust their information need to fulfil their information search goal (Borlund and Dreier, 2014).  The consequential information search leads individual to sieve out informational cues that are not relevant to the task goal (Pirolli and Card, 1999). For example, online users may not open sponsored advertisement enclosed within the website, regardless of its relevancy to the task. 

During information foraging, the investment of time and effort is critical in meeting task goal productively.  In the absence of time pressure, people rely on their preferences to make decisions (Amichai-Hamburger, 2002).  In weighing the cost and benefit of the information search, individual may lean on familiar sources than unfamiliar ones (Savolainen and Kari, 2004). 

This leads us to the following research questions:

• RQ2: What are the users’ attentional focus on information scent patches that emit strong- and weak-information cues?
• RQ3: What are the users’ interpretation of information credibility?

Evaluating users’ informational cues on the different information scent sites can be examined through eye tracking method.  Past researchers have studied eye gazes on websites (Josephson and Holmes, 2002) and task performance on web portal page using eye tracking methods (Goldberg, Stimson, Lewenstein, Scott, and Wichansky, 2002). Other information researchers (Pan, Hembrooke, Gay et al, 2004) found that female users engaged in a shorter mean fixation durations than male users on web pages, and they tend to spend longer time gazing the first page than the subsequent pages. 

Pan et al. (2007) investigated cue from the links on google search ranking results. These studies generally focused on the gender and google ranking page without defining the features and characteristics of different scent on information patches that may influence information searching behaviour. This paper aims to investigate if there are differences in the way users focus on the cues on the different information patches that possess strong- and weak information scent.

Method

Fifty-four students were recruited from Nanyang Technological University, Singapore, in an experiment conducted from February to April 2015. Two of the 54 participants were removed from the eye-tracking data due to technical problem, whereby their eye-movements data were not captured on the selected webpages.  Fifty-two participants (24 male, 46% and 28 female, 54%); aged between 18 – 35 years old (18 – 24 years, 98%,  31-35 years, 2%) were offered the options to choose either one course credit or an honorarium of $15 per hour for this study. 

University students were recruited as they are familiar with the use of internet for information searching for their school work and projects. Information search on the Web is commonly used in schools and universities, employing popular search engine such as Google (Rieger, 2009).

The frequency of their use of the search engines and the use of direct website address URL were reflected in their responses to the questions based on a Likert scale of 1-5, ‘1’ being least frequently to ‘5’, being most frequently: ‘I browse the Internet using search engines for general information search’,  (90.7%),  M=4.50, SD =.84, and  ‘I use direct URL address (website address) for information search.’ (42.6%), M=3.17, SD=1.20.

Informed consent forms were given, read and signed by all participants before the start of the experimental study, in accordance with the university’s Institutional Review Board (IRB) regulations.

Procedure and Stimulus

A quasi-experimental study using the mixed factorial design was employed for this research.  Quantitative and qualitative methods were applied for analysis. Participants assigned for the between-subjects groups performed the browsing and specifically-defined instructions for information search on cuisines and travel-related information.  For the within-subjects, participants carried out two sets of information search tasks for the cuisine and travel-related topics. Post-experiment interviews and surveys were carried out to gain an in-depth understanding of the participants’ cognitive search behaviour. In this study, non-specifically defined browsing-task is task that provides general instructions (see Table 1).  Specifically-defined task is task given additional instructions, e.g., search for ‘seafood restaurants in the east zone’, and for travel site, ‘three places of interest in South Africa’.  Participants were not informed on the task types.

For both browsing and specifically-defined instructions, on food and travel topics, participants were informed that they could independently conduct the search queries, and were given the autonomy to navigate to any website to retrieve the information. In addition, participants were requested to visit two specific websites as part of their information retrieval process. The selected websites were LonelyPlanet.com site for the travel-related information and the food blogger (Food Blog) site for the cuisine-related information (see Figure 1a and 1b). The name of the food blog site is not revealed in this paper to protect the privacy of the personal blogger.

Data collection was gathered from a combination of log-analysis from Tobbi T60 eye-tracking system, interviews and questionnaire surveys. Eye-tracking analysis allows the researcher to generate quantitative data. A common computation of eye tracking metrics includes fixation durations, fixation counts (for scanpath or gaze interactions), mouse-clicks and areas of interest (AOIs). AOIs allow particular elements on the webpage (e.g. headers or headline captions, advertisements, images, text) to be selected for analysis of gaze fixations. A minimum threshold of 100 millisecond (1 second) was set for the eye-tracking metrics, a standard fixation threshold value for eye-tracking studies (see Goldberg and Koval, 1999; Jacob and Karn, 2003; Coltekin, Fabirkant and Lacayo, 2010). 

Fixations refer to stable eye position within a certain threshold of dispersion (Pan et al., 2007). Saccades or scanpath (refers to gaze interactions in this paper) are ballistic movements or sequence of fixations (Pan et al., 2007). It varies in amplitude (size) from small fixated movements to large movements made while gazing the target or when reading (Purvers, Augustine and Fitzpatrick, 2001).

The two websites were chosen for their likelihood to generate a variety of information scent patches, its domain differences, that is, institution and author’s popularity, the contents and design of the websites. Prior to the actual experiment, five participants were recruited for a pilot test of the information retrieval on the travel and cuisine topics. Subsequently, these two sites, Lonely Planet and Food Blog websites were selected as both sites possess a variety of information patches, in which they were deemed appropriate for this research.

Drawing from the theories of source credibility (see Wilson and Sherrell, 1993; Sundar, 1999; Winter and Kramer, 2012),  a priori category was considered for information patches, that is, LonelyPlanet and Food Blog sites, which provide strong information cues (Lonely Planet) and weak information cues (Food Blot).  Lonely Planet being a popular travel publisher, with print over 120 million books (Lonelyplanet, 2016) projects a strong information scent patch for its authoritative and established institution. The little-known food blogger, Food Blog, is categorised as weak information scent patch for the source-unfamiliarity with most users.  The data from the participant’s perception of informational cues on these two websites were generated from the quantitative and qualitative content analysis of the post experiment interviews. Participants were not given any details or opinion on these two websites.

Findings

Following the experiment, participants were asked to rate on the scale of 1–7, 1 being least information credibility, and 7 being the most on information credibility of Lonely Planet travel and Food Blog food blogger websites that they have visited.  The credibility measurement was based on Sundar, Knoblock-Westerwick and Hastall (2007) questionnaire, and modified from a 10-point Likert scale to a 7-point scale. The higher scale set options is likely to produce diminishing returns (Nunnally, 1978; Sauro and Dumas, 2009). As the questionnaire was given as post-task survey, the prevalent 7-point scale set is sufficient to achieve reliability, and this scale makes it easy for participants to respond to the questions (Sauro and Dumas, 2009).

Hypothesis 1a and 1b postulate that users will evaluate information patch with cues that emit strong information scent a high level of information credibility when conducting search on this site. Conversely, users will evaluate information patch with cues that emit weak information scent a low level of information credibility.

Hypothesis 2a and 2b propose that users’ perception of website with strong information scent may spend a longer duration of time (attentional focus) on the information patch, and users’ perception of weak-information scent website may spend a shorter duration of time on the information patch. 

The findings from the paired comparison t-test reflected that there was a significant assessment of the level of credibility in the travel and food blog sites (N=52). For Lonely Planet site,  M = 5.13, SD = 1.51, t(14.99), p < .001 and Food Blog site, M = 3.77, SD = 1.37, t(2.15), p < .001.

Task and credibility assessment

RQ1: Do types of task influence the users’ perception of credibility?

To understand if there was an association between the types of task and the credibility assessment, the factorial analysis was conducted. The data showed no significant main effect on the types of task, browse and specific on Lonely Planet, F(50, 1) = 1.73, p = .075 and Food Blog, F (50, 1) = 2.77, p = 1.50 and the credibility assessment.  The descriptive statistics indicated that the mean credibility assessment of Lonely Planet in relation to both types of task did not project a wide difference, for browsing and specifically-defined task (N = 52), browsing task,  M = 5.08, SD = 1.57, and specific task, M = 5.19, SD = 1.47. 

The mean credibility assessment of Food Blog site to task types revealed that participants of the browsing task group assessed the credibility of the Food Blog site higher (M = 4.00, SD = 1.06) than the specific task group (M = 3.54, SD = 1.61). This may be attributed to the precise search instructions administered in the specifically-defined task that led participants to hold certain expectations of information scents exhibited in the webpages. 

We further examined if task difficulty has effect on the credibility judgement.  The question for task difficulty of the information retrieval of both websites was elicited from the participants after the experiment.  Participants responded to the question using the scale of 1 to 7, 1 being least difficult to 7 being most difficult (reversed coding applied for clearer analysis).

According to correlation coefficient findings, there appeared to be no correlation between task difficulty and perceived credibility for both travel- and food-related search topics – travel search task, r(50) = .038, p = .789, and food search task, r(50) = .255, p = .068. 

Attentional focus and credibility assessment

To examine if there was an effect between the participants’ perceived credibility and their attentional focus (fixations on Lonely Planet and Food Blog sites) on the websites, a 2 x 2 ANOVA was conducted to assess the credibility perceptions on Lonely Planet and Food Blog sites and attentional focus on both sites. Attentional focus (Blackmon, 2012) in this paper is operationalised as the duration of eye fixations and gaze fixations, that is, the scanpath or saccade, on the information patch (webpage).  

The findings indicated that the main effect of the users’ credibility perception and the attentional focus were significant on both websites, F(1, 50) = 3.012, p <.005, partial η² = .204. 

A follow-up test indicated that there was no significant main effect on credibility perception and Lonely Planet site, F(1, 50) = .075, p = .786, partial η² = .001.  The main effect for credibility perception and Food Blog site was significant, F(1, 50) = 11.47, p = .001, partial η² = .187. 

This implication suggested that the perceived level of credibility of the Lonely Planet site has no influence on the attentional focus (duration of fixations) on the travel site, while the credibility perception of the Food Blog site may contribute to the attentional focus on the food blog site. 

Task and attentional focus

We followed up with the examination of the nature of the task instructions on the attentional focus of the information patches.

Browsing task: The duration of fixations on the Food Blog site is longer in the case of the browsing group (M=1.94, SD = 1.27) compared to the browsing group of search task on Lonely Planet site (M = 1.50, SD = 1.00). 

Specific task: On the other hand, in the specific task group, participants spent a longer duration on the Lonely Planet site (M = 1.40, SD = 1.02) compared to the specific task group conducting information search on Food Blog site (M = 1.03, SD = 0.49)

Hence, Hypotheses 1a, 1b and 2a and 2b were partially supported.

Attentional focus and information patches

RQ2 aims to address the user’s attentional focus on information scent patches that emit strong- and weak-information cues. 

The duration of fixations were defined within the areas of interest (AOIs) region to examine users’ gaze interactions. Link-clicks from the eye-tracker logs analysis will be reported under ‘Information scent and cues’ section.

Information Patch and Cues

A 2 x 5 ANOVA was conducted to examine the effects of the information patches (travel and food websites) and informational cues (cues identified as text, pictorial, header, categories and advert). 52 participants of both browse and specific task groups conducted search on a total of 104 webpages. 

There was a significant main effect of strong- and weak-cues patches, F(1, 102) = 28.64, p <.001, partial η² = .594, and the attentional focus (fixations) of cues on the websites.

Text and pictorial cues

A follow-up pairwise comparisons reflected that participants’ fixations on the pictorial cues on the Lonely Planet webpages was significant, compared to the Food Blog webpages, F (1,102) = 42.96, p <.001, partial η² = .296.  The pictures displayed on the travel webpages are embedded with hyperlinks and captions that garnered attention to view the elements, and enticed the users to click on the links. The images on the Food Blog site are static with no links enclosure, and the pictorial elements might have been demonstrated as weak informational cues.

The descriptive statistics of the duration of fixations and the gaze fixations are presented in Table 2.  On both Lonely Planet and Food Blog sites, engagement of text cues appeared to be the salient information scents (Figure 2). On Food Blog site, the text cues are displayed icon-cues to direct users to read information on another page. When participants allocated a shorter time on pictorial cues, the fixation time on the text cues was longer, an indication of focused reading. However, it is unclear if the brown font in small font size posed a problem in the users’ reading of the contents on the food blog site.

Observations on the eye-tracking log showed that on the food blogger site, longer scanpaths were recorded within a short duration time on visual cues, an indication of browsing pattern on the cues.  According to the researchers (Trevarthen, 1968; Graupner, Pannasch, and Veilichkovsky, 2011) a short fixation combining with long saccades indicate a quick evaluation of the spatial layout of a site. A sample of gaze interaction (Figure 3a and 3b) illustrated that with more pictorial cues displayed on a strong information scent patch, gaze interactions tend to be wide. 

Text cues. There was no significant main effect on the both Lonely Planet and Food Blog information patches and text cues, F(1,102) = 7.35, p = .008, partial η² = .067. The fixations and reading text contents demonstrated as top activities on the travel and food websites.  Both sites display text links in different formats. The travel site embeds its text links within the messages and pictures, and may elicit a strong scent for users to engage in these elements. On the other hand, on the food blog site, hyperlinks ‘Read on’ and ‘Further post’ text icons, are displayed separately below the message section.  In the absence of hyperlinks on the pictures, users were drawn to click on these text links on the Food Blog site.

Header cues. There was no significant effect on headers cue and information patch, F(1,102) = .61, p = .438, partial η² = .006. The headline captions are available on both sites, and users appear to use the headlines links for both tasks.

Categories and advertisement cues – cluster effect

Categorical cues. The main effect for categorical links fixations was significant, F (1,102) = 16.27, p <.001, partial η² = .138. A list of categorical links of food classifications is displayed at the bottom of the food blog site, and there was scant categorical links on the travel site.

The categorical links and an advertisement banners are located at the bottom of the Lonely Planet webpage (Figure 4a).  Participants seem to adopt a symmetrical pattern in the viewing the contents.  Both categorical links and advertisements were viewed fixated in tandem. 12 participants (23%, N = 52) fixated on the categorical links, and only 1 participant who viewed the categorical link also fixated on the advertisement banner (see sample in Figure 4a and 4b).  

Advertisement cues. The main effect for both travel and food blogger sites was significant, F(1,102) = 21.65, p = < .001, partial η² = .309. A high number of advertisements are displayed on the right panel of the food blog site. The low fixations and scant number of clicks on the advertisement of both travel and food blog sites denote weak information scent of the visual elements.

40% had zero fixation on the advertisements displayed on the top right panel of the webpage.  48% of the participants did not view both the advertisements and categories link that are located linearly at the bottom of the advertisements (Figure 4a and 4b). 

The areas of interest illustrated a clustering fixation effect in users’ foraging behaviour, demonstrating that participants scanned elements as a whole and associating them with the types of cues they had fixated on earlier. In this site, the inverted-L section (Figure 4b - blue box within the AOI region) emits no scent.  Generally there was absence of fixations in the section, eventhough there are systematic categorical links situated below the advertisement banners.

Information scent and cues

Information scent (scent-following) is operationalised as the number of clicks on the hyperlinked contents and visual elements such as images, icons, headline captions (header), to assist users to move to other webpages.

The link-clicks data indicated that the elements of pictorial, text and headers hypertextual links emerged as strong information scents in Lonely Planet site, and directional-cue (i.e. ‘read on’, ‘older post’ links) and categories links emitted strong information scents in Food Blog site.

Averagely, participants navigated to a total mean of 11 pages and 7 pages to the travel site and the food blogger, respectively. In browsing task group, participants engaged in higher link selections in travel webpages as compared to food blog webpages (Figure 5b). In the specific task group, there was no wide difference reflected in the links engagement for both sites, and the scent-following was lower in this task group as compared to the browsing group. This phenomenon may be attributed the effort expensed by users to acquire precise information as required by the specific task instructions, that led them to incur higher fixation durations in the information patches, resulting in lower navigation patterns.

Users’ perception of information credibility – qualitative analysis

A qualitative assessment from the interview data was conducted to address RQ3 in examining the users’ interpretation of the characteristics of information credibility. 

Through a one-cycle thematic coding approach, participants’ comments were categorised into key credibility interpretations. Saldana (2013, p. 60) suggested that depending on the nature and goals of the study, researcher may find that one coding method alone will suffice. 

Strong information scent site

Users who perceived the website with strong information scent relate the source of website as credible. The information scent is derived from cues offered by the expertise knowledge from the industry, official institution and reviews by others.

Familiarity of the source is another element where participants gauge the information credibility. As a participant (P047) stated that ‘I will look at the source providing the information – if they are from the industry.’, P050 put it that information needs to be from the ‘Government website and it’s peer-reviewed’, and participant P027 articulated that ‘LonelyPlanet is credible, people know the website for travel information, from book to website.’

Well organised information with combination of in-depth information and pictorial enhancement were seen as a professionally designed website.  Organised contents that are categorically arranged contribute to the believability of the information. One participant commented on Lonely Planet’s site ‘The website looks credible. LonelyPlanet is professionally designed. There are detailed text and images’ (P035), ‘Travel site is authentic. LonelyPlanet is credible’ (P045). Other users attributed ‘content, descriptions, pictures’ (P002), ‘categories, and in-depth information for each section’ (P017) as significant. Some participants preferred to see ‘detailed information supported with images’ (P010), and ‘professional written language used’ (P008).

Informational cues such as updated information and consistent positive reviews from other readers were viewed as credible.  As participant (P054) shared that ‘It is a place I would start searching because it is established and reliable, and quite up-to-date.’ 

Consistent positive comments and opinions from other reviewers appeared to emit a strong informational cue leading the users to attach credibility to the site, as shared by this participant: ‘On comparison of reviews – most readers share the same reviews’ (P022), ‘It is trustworthy and an official websites, with the reviews’ (P008).

These reviews should also provide an unbiased opinion, as a participant commented that ‘balanced-viewpoints and present (presentation) of the pros and cons of the subject’ (P049).

Weak information scent site

On website with perceived weak information scent, users attributed to lack of images, cluttered and disorganised interface, poor font, and saturation of advertisements: ‘It is not something I want to search because of more text.  In xxx (another well-known food blogger, name removed for privacy) food blogger, there are more images and usually more appealing, in HD version’ (P018), ‘The texts (referring to Food Blog site) are cluttered’ (P049).

Users relate pictorial contents to deriving a sense of gratifying experience while searching the webpages. ‘It gives me ‘whole experience’, and includes pictorial evidence’ (P051).

Poor design and interface of the website projects a weak information scent cue, as articulated by these participants: ‘The website is not well designed.  Information on the website is not elegant’ (P009), ‘It is not organised’ (P030), ‘I don’t like the font, it is hard to read’ (P050), ‘The page is messy, (it) doesn’t categorise the type of food.  I can’t find what I want, there are no categories’ (P045), ‘There are too many advertisements’ (P044).

Participants’ assignment of trust to the websites may be attributed to the familiarity of the source.

The participants’ implicit level of expectation and the absence of prior information search on the website led to them interpreting the patch as weak informational cue: ‘I am not familiar with this site, the blog is not as trustworthy’ (P019). Similarly, another user articulated that ‘he is not familiar with the source and it is not as trustworthy’ (P022).

Discussion

This study offers insight on the interplay of information credibility, information scent and attentional focus on cues during information foraging activities. The results demonstrated that the users assigned credibility on the information patches (websites) that elicit strong information cues. In this research, the characteristics and conditions of the cues conveyed by the information scents have influential effects on the perceived information credibility of the sites.   

Source characteristic constitutes peripheral cues that may determine individuals’ mental effort (Winter and Kramer, 2012) in foraging information online. The differences in the LonelyPlanet and the food blogger (Food Blog) websites, rest on their level of popularity, the language, aesthetical designs (font and layout) and the amount of advertisements displayed on the sites. During information retrieval, the evaluative faculty of judging the credibility of the information scents on the webpages may overwrite the prior knowledge of source credibility. Common expressions garnered from the post-experiment interviews revealed that clean interface, website reliability, reviews, professional language used and the accessibility of hypertext links were some favourable elements preferred by the users to assist in their information retrieval and web navigation.  

The condition of tasks appeared to have some effects on the link selections of the information patches. The browsing task group navigating the perceived strong-scent travel site engaged in a higher link selections than the specific task groups.  It was also observed that the specific groups of both sites recorded lower navigations compared to the browsing group (amid mild difference for food blog site).  This is in part due to the demand of the task instructions that led the participants to incur higher attentional focus on viewing and reading to look for precise information. For food blog site, both the browsing and specific groups reported a closely equal amount of link engagements. The differential in the scent-trail following, that is through link clicks, is likely attributed to the higher number of relevant hyperlinks located within the travel webpage than on the food blog webpage.

The responses suggested that participants perceived the strong information scent patch as reliable, and weak scent patch as untrustworthy.  The results reflected that users navigated to more webpages on the strong scent patch (Lonely Planet) as compared to the weak scent patch (Food Blog) webpages. This is attributed to the salient cues such caption-links are attached to the pictures, and text link are integrated in the messages to assist information seekers to navigate to the relevant information sources swiftly. The salient link mechanisms exhibit a sense of credibility of the information scents on the information patches. Information patches with relevant contents increase users’ perception of a positive evaluation of the website (Kalyanaraman and Sundar, 2006), and capture their attention and they are likely to ‘linger’ on the website (Little, 2001, p. 53).  In contrary, the unfavourable features of the low information-scent patch, Food Blog, as perceived by the users, are attributed to poor design, font colour and size and the layout of the site. Comments such as cluttered and disorganised contents, unfamiliar source, untrustworthy, and display of too many advertisements, were some responses. From the perspective of the users, the high number of advertisement banners and videos, located on the right panel of the website, was interpreted as irrelevant cues and exhibit as weak information scents. 

Conversely, we found that the nature of task instructions appeared to have some impact on the users’ cognitive evaluation and credibility assessment as they held certain level of expectations of the information scents on the websites.  Although perceived task difficulty did not appear to be associated with the credibility perception, the participants from the specifically-defined task group, as compared to the general browsing task group, evaluated the level of credibility lower on the weak information scent patch (food blog site). This information behaviour sheds light on the users’ evaluation of information scents and credibility of the information patches, such that when the task goal is specifically defined, the expectation of the level of credibility of the information patches is higher. 

In task focused information search, users may tap on their evaluative faculty, guided by the cumulative effect of information scent (see Sundar et al., 2007).  Intrinsic judgement based on prior experience, topical knowledge, or popularity of the author and institution, assist users in weighing the cost and benefits of information foraging.  Drawing from the feedback of the participants, they leveraged on their cognitive appraisal of the emergent cues exhibited in the information patches. Users related to information credibility with the accessibility of reviewers’ opinions implies a heuristic information processing strategy, to leverage on others’ knowledge and experience. Information seekers adopt search strategies (Joho, Jatowt and Blanco, 2015) and cognitive search intent (Kato, Yamamoto, Ohshima, and Tanka, 2014) during information processing.  Further study will be useful in understanding how task specificity affects the level of heuristic evaluations on information patch during information processing.

Following the investigation on the attentional allocation of users in the different information patches, it was observed that within a perceived higher credible information patch, there exists weak information scent where no gaze interaction or reading patterns was present. Engaging in text links and contents were the top activities of information processing on both travel and food blog sites.  The accessibility of the hypertextual links on both information patches contributed to the information scent interactions.  Previous eye movements studies suggested that text were usually the first activity online readers engage in news (Josephson and Holmes, 2002), and print advertisement (Rayner, Rottello, Stewart, Keir, and Duffy, 2001). Eye-tracking researchers expressed that the duration of fixations and the extent of the saccadic gaze are related (Tatler and Vincent, 2008).

Although the qualitative content analysis showed that participants associated credibility with the display of quality pictorial cues (images), they may not necessary interact with these cues in strong information scent patch. This implies that amid users’ desire for visual or animated cues as offering them gratifying experience during web navigation and search, with task or time considerations, engaging in textual contents is still a main foraging activity on both strong-scent and weak-scent sites. Individuals are able to attend to the appropriate cues and categorizations (Blair et al., 2009) and discriminate between the priorities of search task or salient cues at hand during information foraging. Therefore, an information patch perceived to possess high source credibility may emit weak information scent in the absence of relevant information cues.

Analysis from the eye-tracking logs reported some significant findings in that users engaged cues as a whole rather than as a single element.  Observations from the gaze plots within areas of interest data shed light on the clustering effect of information foraging. For example, users adopted the reduction approach of filtering non-relevant cues, particularly the region with sponsored advertisements. Participants interacted with the cues as whole stimuli rather than view them as a single stimulus.  For example, when there was absence of gaze interactions on the advertisement located at the top right panel, there was no fixation or engagement of the categorical links located below the pictorial cues.  Similarly, users’ fixations at the top region were followed by scanpath present at the bottom region. It is interesting to note that the inverted L-region displayed on a two-fold webpage emitted as weak-scent or no-scent patch where no or scarce link clicks was reported.  Interface design projects as informational cue and the way information and contents are organised with hyperlinks are enacted as information scent.  From this examination of information scent and cues, it was observed that individuals associate attention allocation of the second cue with the first cue is determinant on whether the first cue emits a strong scent. Hence to enhance the categorical links to assist users in navigating the webpages, the hyperlinks would derive higher level of attention if placed in a strategic position together with cues of strong scent.

This study demonstrated that the number of advertisements links enclosed within the site affect the scent cue perception, in particular in the low information scent patch.  The results in this study is in contrary with Kalyanaraman and Ivory (2009)’s studies that a picture attachment to a relevant advertisement generates higher information scent and positive attitude of users. The perception of irrelevancy, harmful (undesirable element of virus attached to the advertisement) and issue of objectivity were some reasons for the non-engagement of this scent cues, a shared by participants. The perceived credibility of the information scent thus influences the engagement of cues.  This finding provides insight for website authors and advertisers to consider enhancing information scent cue involvement rather than cue reduction of users, by presenting the information and cues in a meaningful and an interactive form.

Conclusion and limitations

Overall, this study of information scent cue contributes to the concept of information scent in foraging theory.  The findings showed some evidential support on information scent and cues engagements during information retrieval. Specifically, we incorporated the notion of information credibility, and eye movement analysis in investigating users’ attentional focus and gaze interactions in relation to information scent. The experimental results, using quantitative and qualitative methods, demonstrated affirmatively the empowerment of cues and hypertextual links that shape users’ information search decisions.  Information patch with strong scent garners higher attentional focus on the site, while the site with weak information scent may fail to gratify users’ information search experiences.  The information scents are exhibited in the form of salient headline captions, pictorials, text and categorical links. The observations showed that most users automatically conducted cue reduction strategy in filtering non-relevant elements, and cues position influences gaze interactions and scent engagement.  In foraging patterns, the results demonstrated that there was a clustering pattern in information scanpath behaviour.  Significantly, this has implication in the way cues can be displayed on the webpages to optimise search outcome.

There are limitations in the study that warrant attention. Firstly, the findings from the small sample size may not be appropriate to generalise across broad population.  However, this preliminary study provides an evidence of the heuristic nature of informational and task decision phenomenon in the context of information scent and allocation of attention during information foraging process in online environment.  Task conditions influence the expectation of the information scent and credibility of the information patches. Secondly, research in this area can be extended to investigate the variables of prior topical knowledge, information literacy, motivations of task goal and complexity of the task relative to the engagement of information scent. Thirdly, it would be beneficial to examine the types of overt and covert information scents in influencing heuristic information behaviour.

Funding

This research has received funding from Nanyang Technological University, Wee Kim Wee School of Communication and Information, Singapore.

Acknowledgements

The authors would like to thank the two anonymous reviewers for their insightful and valuable feedback that contributed to the improvement of this paper.

About the authors

Helena Lee is a doctoral student in the Wee Kim Wee School of Communication and Information, Nanyang Technological University, Singapore.  She received her MSc in Knowledge Management from Nanyang Technological University.  Her research interests include information foraging behaviour, social media and eye-tracking studies.  She can be contacted at: lees0143@e.ntu.edu.sg.
Natalie Pang is an Assistant Professor in the Wee Kim Wee School of Communication and Information, Nanyang Technological University. Prior to joining NTU, she has worked on public opinion research in The Gallup Organization, citizen science and participatory methods in Monash University and Museum Victoria. She specializes in the area of social and community informatics, examining the use and impacts of social media of communities in crises, social movements, and everyday life.