How Users Search the Mobile Web: A Model for Understanding the Impact of Motivation and Context on Search Behaviors
Article Category: Research Paper
Published Online: Sep 01, 2017
Page range: 98 - 122
Received: Jan 03, 2016
Accepted: Mar 03, 2016
DOI: https://doi.org/10.20309/jdis.201608
Keywords
© 2016 Dan Wu, Man Zhu, Aihua Ran
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
With the development of the mobile Internet, mobile search has become an important means for people to access information. According to the
As the popularity of mobile search among young people continues to grow, many mobile service providers have engaged with mobile search applications development to attract college students. In order to design applications and services for particular information needs and mobile contexts, it is necessary to understand how user motivation and context influence their search behaviors. To this end, we conducted an online diary study with college students. Our research questions are:
RQ 1: How do mobile search motivations influence mobile search behaviors? Do mobile search sessions driven by different motivations relate to each other? RQ 2: How does mobile search context influence mobile search behaviors? What is the relationship between different context dimensions?
Based on the investigation results we developed a mobile search behavior model to understand the relations between different variables.
Malhotra, Galletta, and Kirsch (2008) formed a theoretical basis for discerning whether user behaviors resulted from perceived external influences or personal volition. Since then, user motivation has been investigated as an important influencing factor in many information-seeking behavior studies. Chang and Yang (2009) applied motivation theory to participating behaviors, Bilandzic and Foth (2013) to social learning behavior, and Kim and Adler (2015) to online information sharing behavior. However, these studies did not take into account the device where mobile search behaviors happened, omitting comparisons between mobile and desktop search.
Recent studies investigate how user motivation affects information behavior in the mobile context. Ciampa (2014) examined whether and how Malone and Lepper’s taxonomy of intrinsic motivations for learning applied equally well to mobile learning, and Huizenga et al. (2009) applied motivation theory to mobile game-based learning. Church and Oliver (2011) studied many aspects of mobile search, and proposed a motivation classification theory to help understand why users use mobile devices to search for information. They treated search motivations as a part of mobile search behaviors, rather than a driving factor of mobile search activity, and indicated that more effort is needed to study the relationship between mobile search and information-seeking motivation.
Search context was identified as one factor that plays a vital role in mobile search behavior. Kong et al. (2015) proposed two types of short-term context: pre-search context and in-search context, and suggested that pre-search context triggers the search. Liu and Wu (2015) suggested that contextual information, which included time, weather, and emotions, influences a user’s preference on a certain point of interest. Taking geographical place and user behavior as contextual information, Kiseleva (2015) studied their influence on searching and browsing behavior. Gasimov, Magagna, and Sutanto (2010) divided contextual information into device information, user information, and environment information, and presented a simple architecture for adaptive mobile Web page browsing.
Recent studies discuss the effect of mobile devices on mobile search behavior. Church and Oliver (2011) suggested that limitations in screen size and the type of interactions and input that mobile devices can support play an important role in shaping the way that mobile users use mobile search engines. Kamvar and Baluja (2006) found out that Google’s PDA logs related to searches originating on PDAlike devices with sophisticated input capabilities. Park and Ohm (2014) introduced an integrated research model to examine user acceptance of mobile map services in mobile devices. Studies of mobile devices mostly focus on screen size and how it works with search behavior, such as map search behavior. Device is also considered one of multiple contextual dimensions.
Most current research focuses on specific context, however, such as studies on place, or other contextual influences on mobile search, and lacks a systematic analysis of mobile search context. In this paper, we analyze college students’ mobile search behaviors in a multiple-dimensional context, and build a model of mobile search behaviors based on context.
We conducted an online diary study, in which 30 participants agreed to share daily mobile search records on their phone logs with our researchers. The experiment lasted 15 days, and the participants were asked to do mobile search as usual in their daily life. An application usage-monitoring tool, based on SQLite Developer, was installed on participants’ phones to record their mobile search behavior details.
Online questionnaires were distributed to a wide range of college students to explain our experiment and investigate the frequency of mobile search in their everyday life, the number and type of mobile applications installed in their phones, and their level of mobile search ability. We used these questionnaires to analyze the background information of possible participants. We decided to select skilled users to participate in our experiment for two reasons. One reason was that we needed more mobile search records to do log analysis while the other was that experienced users were typical, and had much to say about mobile search behaviors. Participants were identified among the finalists of a search contest hosted by Baidu, China’s largest search engine. Electronic questionnaires were sent to their e-mails, and 58 students wrote in reply. We recruited 30 participants among these respondents from all over China.
The participants are all college students (9 male and 21 female), including 21 undergraduates and 9 postgraduates with average age of 21 years old. They come from seven universities with different disciplinary backgrounds such as computer science, information science, library science, linguistics, finance, economics, psychology, and surveying and mapping.
By installing AWARE, an Android framework for logging and sharing mobile context information, on participants’ smartphones, we were able to collect logs of daily phone use. After signing our participant agreement, participants were asked to promise that they would do mobile search at least once a day. Each of the three laboratory personnel engaged in the experiment managed ten participants. They checked the log records every day to ensure that there was no discontinuity of the AWARE framework and that each participant performed mobile search at least once a day.
Before the experiment, there were short periods of trial operation to test AWARE’s function and output. System operation, application usage, keyboard input, and location information were recorded in the experiment. Voice input, accounts and passwords, information received, and browsing history were not recorded. Since AWARE only recorded keyboard input, which participants could control, there was no privacy issue. Because we told the participants what kind of data we collected, they were reassured regarding the privacy of their data.
Structured diaries were distributed to participants by laboratory personnel every day in which they were to identify the most impressive search each day. Information needs, information motivations, search sites, and context were requested in the structured dairy, as well as any interruption of the mobile search session. By collecting the diary records from user phone log files with SQLite Developer, a database administration tool, deeper analysis about users’ mobile search behaviors was made possible. Any irregularities or variations in searches were explored in semi-structured interviews with participants.
In this paper, we define mobile search context as the personal attributes and surroundings of the participants when doing mobile search, including the physical environment (e.g. indoor, outdoor, and specific place) and the social environment (e.g. alone or with others). There are 12 defined dimensions of participants’ mobile search context (Table 1), which can be divided into three categories. We get this information from the structured diary and the phone log, and do correlation analysis or difference analysis between context dimensions and mobile search behaviors. The analysis tool we use is Statistical Product and Service Solutions (SPSS 19.0), and analytical methods are cross tabulation and one-way analysis of variance (ANOVA).
Dimensions of participants’ mobile search context.Category Dimension Definition Personal attributes of the Gender Sex (male or female) participant Age Participant age Grade Freshman, sophomore, junior or senior Major Academic field Physical environment of the Time Year-month-day-hour-minute-second participant Place Indoor or outdoor; specific location Weather Summer, rainy or sunny Device Phone type and screen size Social environment of the Activity Event involved while searching: resting, working, participant studying, shopping, traveling, etc. Relations Accompanied or alone Importance Whether the search task is urgent Portal Application used
We explored four aspects of mobile search behaviors: search sessions, search queries, information needs, and emotion changes as follows.
A
Participants were asked in the structured diaries to identify which motivations defined their most impressive search sessions for each day. We found that it was common for users driven by multiple motivations to finish one mobile search session. We define six motivation categories in our research to meet every search session.
Table 2 is a matrix of participants’ information motivations. It illustrates searches with one or two motivations. For searches with more than two motivations, we add one for each pair of motivations in that particular search. For example, one participant chose time killing, curiosity, and social relations as the motivations for one mobile search session. We add one to the frequency for each of the following: the relationship between curiosity and social relations, the relationship between time killing and curiosity, and the relationship between time killing and social relations. After all the integers of frequency are added, we calculate the percentages of information motivations by Equation (1) (
Matrix of search motivations.Information Curiosity Time killing Knowledge Life service Social relations Others Total Curiosity 79 (17.56%) 38 39 14 3 4 177 Time killing 8.44% 31(6.89%) 22 8 2 5 106 Knowledge 8.67% 4.89% 88(19.56%) 16 1 7 173 Life service 3.11% 1.78% 3.56% 95 (21.11%) - 6 137 Social relations 0.67% 0.44% 0.22% - 10(2.22%) 1 17 Others 0.89% 1.11% 1.56% 1.33% 0.22% 19(4.22%) 40 Total 39.33% 23.56% 38.44% 30.44% 3.78% 8.89% 650 (144.00%)
Results show that 71.56% of all mobile search sessions came from a single search motivation, and 24.44% came from two search motivations. The remaining 4% came from three search motivations. Mobile search sessions were mainly motivated by curiosity and knowledge, which accounted for 77.78% of the total. Search sessions motivated by daily life service and time killing were the second largest group. Even though there was a low rate of searches driven by building or avoiding social relations, we concluded that participants were driven by relatively balanced motivations to perform mobile searches in their daily life, and they relied on mobile search to meet various information needs.
On the other hand, every mobile search session was driven by 1.44 types of motivations on average. Search sessions driven by curiosity, time killing, and knowledge shared high rates of correlation. More specifically, time-killing searches were prone to be influenced by other motivations such as curiosity and knowledge. Search sessions for life service were relatively independent of others.
We explore query frequency and repetition pattern to probe the motivation behind mobile search sessions. We add up the queries generated by different types of search motivations. Each search session is presented by a typical query, which means that no matter how many queries appear in a search session, we count only one query here.
Motivations and query frequency.Motivation Query frequency (%) Curiosity 180 (40.00%) Time killing 94 (20.89%) Knowledge 143 (31.78%) Life service 128 (28.44%) Social relations 20 (4.45%) Others 33 (7.33%)
Query in different motivations.Motivation RT≤2(%) RT>2 (%) Total (%) Curiosity 39 (14.83%) 9 (3.42%) 48 (18.25%) Life service 70 (26.62%) 40 (15.21%) 110 (41.83%) Time killing 26 (9.89%) 19 (7.22%) 45 (17.11%) Knowledge 40 (15.21%) 17 (6.46%) 57 (21.67%) Others 3 (1.14%) - 3 (1.14%) Total 178 (67.68%) 85 (32.32%) 263 (100%)
The information need types are categorized in reference to Wikipedia
Categories of information need types.Category Examples Natural science Chemistry, physics, astronomy Social science Economics, history, celebrities Culture & art Museum, exhibition, performing arts, photography City & architecture Chicago, Wuhan, Roman catholic cathedral Education School, courses, studying abroad Computer skills Software and hardware, operating skills News & events Social news, political news, gossips Medical treatment & health Hospitals, diseases, medicare Recreation Music and videos, restaurants, accommodation Transportation Roadmap, train and airplane routes Social networking QQ, WeChat, discuz!, net community Daily shopping Clothing, books, food, electronics Others Religions, weather, sports
We analyzed the motivation and information need types via cross tabulation and found significant correlations between them, shown in Table 6. We also analyzed the three significant correlations to understand the information need types under different motivations, as shown in Table 7.
Search motivation and information need type. < 0.05 < 0.05Motivation Correlations (p-value) Curiosity 0.038 Time killing 0.118 Knowledge 0.020 Life service 0.000 Social relations 0.358 Social avoidance 0.910 Others 0.354
Top three relationships between search motivation and information need type.Motivation Top 1 (%) Top 2 (%) Top 3 (%) Curiosity Culture & art (31.1) Social science (22.2) Recreation (6.7) Knowledge Culture & art (22.4) Social science (16.1) Computer skills (12.6) Life service Culture & art (18.0) Recreation (15.6) Daily shopping (12.5)
Tables 6 and 7 indicate that three types of motivations had significant correlations with information needs. Driven by different search motivations, participants appear to have diverse information need types. We concluded that the college students’ needs for culture and art information were commonly higher than for other information types. When participants searched to satisfy their curiosity, their information needs for social sciences and recreation were relatively higher. When they searched to increase their knowledge, their information needs for social sciences and computers were also relatively higher. When participants decided to search to satisfy daily life needs, they tended to find information on recreation and shopping.
To understand the different emotion changes under various motivations, we performed a cross tabulation analysis between search motivations and emotion changes, as shown in Table 8.
Search motivation and emotion change.Motivation Correlations ( Curiosity 0.444 Time killing 0.071 Knowledge 0.075 Life service 0.121 Social relations 0.538 Social avoidance 0.847 Others 0.000
Table 8 shows that search motivation and emotion change did not correlate significantly with each other except for “others” which includes motivations of finishing homework, finding discount information, etc. For example, when users tried to search for discounts or to finish homework, they generally tended to reflect a better mood.
We obtained query frequency data from the structured diaries and the search logs, and counted the query frequency at different periods. Table 9 shows the correlation analysis results, in which we saw that only time and place significantly correlated with query frequency.
Search query and context.Category Mobile search context dimensions Correlations ( Personal attributes Gender 0.351 of the participant Age 0.271 Grade 0.577 Major 0.326 Physical environment Time 0.000** around the participant Place 0.000** Weather 0.699 Device 0.154 Social environment Activity 0.190 around the participant Relations 0.280 Importance 0.214 Portal 0.881
We divided one 24-hour day into four periods—00:00–06:00, 06:00–12:00, 12:00 –18:00, and 18:00–24:00. We chose this time division to make the length of the four periods equal and make 00:00 the beginning of a day. As mobile search is convenient and instant and the users may do it at any time, we did not take mealtime into consideration. The time division we used might not be optimal as it cut through mealtime, which probably had some influence on the analysis.
Place here contained six specific locations: home or dormitory, workplace, study place, in public, in transit, and on vacation. In order to understand the query frequency in four periods and six locations, we gave two cross tabulations as shown in Tables 10 and 11.
Top two relationships between time and query frequency.Time Query frequency (%) 18:00–24:00 44.2 12:00–18:00 40.0
Top two relationships between place and query frequency.Place Query frequency (%) Home or dormitory 54.7 Study place 28.7
Table 10 indicates that about 44.2% of the queries appeared between 18:00 and 24:00, and 40% between 12:00 and 18:00. Furthermore, the place and query frequency had significant correlations. About 54.7% of the queries were formulated at home or in the dormitory, and 28.7% in a study place. This might have something to do with the college students’ lifestyle—they often stayed up late and were active in the afternoon, while dormitory and study place were two main places of their college life.
The information need and the mobile search context are both parts of information search behaviors. The information need is expected to have something to do with the various attributes of the participants and their surroundings. The correlation analysis results are shown in Table 12. If the users tend to search for some information via various portals at the same time, we make analyses of the information need and every portal.
Information need and mobile search context. Note. * < 0.05, “A” stands for all correlated, which means all subdivisions of a certain dimension correlated with another dimension or a subdivision of the dimension.Category Mobile search context dimensions Correlations (p-value) Personal attributes of the Gender 0.000 participant Age 0.325 Grade 0.139 Major 0.314 Physical environment of the Time 0.232 participant Place - Weather 0.412 Device 0.253 Social environment of the Activity 0.002 participant Relations 0.577 Importance 0.032* Portal A
The social environment of the participants had significant correlations with their information needs, while their physical environment did not. To get a better understanding of the correlations, we made the following additional analyses.
Top three relationships between information need and gender.Gender Top 1 (%) Top 2 (%) Top 3 (%) Man Culture & art (23.0) Social science (14.1) Natural science (14.1) Woman Culture & art (24.8) Social science (18.7) Recreation (10.2)
Table 13 shows that the men and women were mainly concerned about culture and art information. In addition, the women cared more about social sciences and recreation information, whereas the men preferred natural science information. This phenomenon shows that the information need differed between men and women, and that women tended to do mobile search when relaxed.
Top three relationships between information need and activity.Activity Top 1 (%) Top 2 (%) Top 3 (%) On break Culture & art (24.4) Social science (18.1) Recreation (8.1) Working Culture & art (40.0) Social science (30.0) Education (15.0) Studying Culture & art (23.1) Natural science (15.7) Social science (13.9) Shopping Recreation (57.1) News & event (14.3) Social science (14.3) Traveling Recreation (28.6) Culture & art (21.4) Social science (14.3) Others Culture & art (22.6) Social science (16.1) Natural science (12.9)
Top three relationships between information need and importance.Task importance Top 1 (%) Top 2 (%) Top 3 (%) Important Social science (24.7) Culture & art (16.7) Recreation (11.4) Neutral Culture & art (27.5) Social science (10.8) Natural science (10.3) Unimportant Social science (28.0) Culture & art (25.8) City & architecture (8.0)
Information need and search portal. < 0.05Search portal Correlations (p-value) Browser 0.000 Video software 0.007 Music software 0.000 Transportation software 0.000 Social software 0.000 Shopping software 0.000 Application store 0.000 Education software 0.023
Top three relationships between search portal and information need types.Search portal Top 1 (%) Top 2 (%) Top 3 (%) Browser Culture & art (23.6) Social science (18.4) Natural science (9.2) Video software Culture & art (63.2) Recreation (10.5) Social science (5.3) Music software Culture & art (100) - - Transportation software Transportation (47.6) Recreation (33.3) Others (14.3) Social software Social science (45.9) Social networking (16.2) Culture & art (13.5) Shopping software Daily shopping (57.1) Recreation (35.7) Social networking (3.6) Application store Computer skills (80.0) Recreation (20.0) - Education software Culture & art (66.7) Natural science (25.0) Others (8.3)
The activity and information need types significantly correlated with each other (Table 14). Participants cared about recreation information when taking a break, education information when working, and natural science information when studying. In addition, participants preferred culture and art information when traveling.
The search task can be important or urgent sometimes. From Table 12, we knew that search task importance significantly correlated with the information need types. When searching for social science—especially economics and celebrities— participants considered it to be important, while culture and art were considered unimportant.
Search portal and information need types were correlated, but the prominent correlation between each search portal and information need types was slightly different, as shown in Tables 16 and 17.
As the above tables show, mobile search portal and the information need had significant correlations. Besides culture, art, and social science, the information searched by browsers was also about natural science, so browsers could serve as a good search portal for participants’ daily studying. The information searched by video application and music application was mostly about culture, art, and recreation, and this conformed to the resources characteristics of the search portal itself. When setting traveling and traffic, social networking, shopping, and learning software as portals, the information need types were consistent with the characteristics of those search portals. When participants were in the application stores, they generally searched computers and recreation applications.
Emotion change in mobile search meant the change of user emotion after searching for information on a mobile device. It is divided into three categories: better, worse, and no change.
From Table 18, we knew that among the 12 dimensions, only gender, task importance, and portal significantly correlated with emotion change. We also saw that the social environment more closely correlated with emotion change than the other two context categories. We examined every search portal with respect to emotion change and learned that all the search portals correlated with emotion change in mobile search to some degree.
Emotion change and context.Category Mobile search context dimensions Correlations ( Personal attributes Gender 0.013* of the participant Age 0.342 Grade 0.451 Major 0.182 Physical environment Time 0.147 of the participant Place - Weather 0.723 Device 0.247 Social environment Activity 0.108 of the participant Relations 0.230 Importance 0.000** Portal P
Table 19 shows that the men’s emotion change was more significant than that of the women. This might have something to do with the information for which they searched. For example, the women tended to search for recreation information, while the men preferred natural science. Men became happier when they got knowledge, while the women focused more on fun.
Relationships between gender and emotion change.Gender Better (%) Worse (%) Neither (%) Man 42.2 10.4 47.4 Woman 28.9 9.2 61.9
Relationships between emotion change and task importance.Task importance Better (%) Worse (%) Neither (%) Important 45.6 17.5 36.8 Neutral 33.8 7.8 58.3 Unimportant 20.5 5.3 74.2
We analyzed the correlation between every search portal and emotion change. It turned out that some of the search portals significantly correlated with emotion change. As seen in Tables 21 and 22, 89% of the participants’ mood did not become worse when searching through a browser. This illustrates the utility of the browser for the participants. When using social software, the participants did not have dramatic changes in emotion.
Emotion change and search portal. < 0.05 < 0.05Search portal Correlations (p-value) Browser 0.032 Video software 0.198 Music software 0.321 Transportation software 0.740 Social software 0.021 Shopping software 0.586 Application store 0.390 Education software 0.052
Relationships between emotion change and search portal.Search portal Better (%) Worse (%) Neither (%) Browser 35.0 11.0 54.0 Social software 13.5 8.1 78.4
The dimensional context of mobile search may have correlations between any two dimensions. We explored every significant correlation to better understand how the mobile search context affected mobile search behaviors. The analysis results were presented in Tables 23–34.
Personal attributes and physical environment. Context Personal attributes ( Physical environment ( Gender Age Grade Major Time Place Weather Device Personal Gender - - - - 0.673 P - - attributes Age - - - - 0.347 0.748 - - Grade - - - - 0.284 0.345 - - Major - - - - 0.529 0.653 - - Physcial Time 0.673 0.347 0.284 0.529 - 0.354 0.736 - environment Place P 0.748 0.345 0.653 0.354 - 0.415 0.866 Weather - - - - 0.736 0.415 - 0.593 Device - - - - - 0.866 0.593 -
Gender and search place. < 0.05At home or dormitory ( In workplace ( In study place ( In public ( On the way ( On vacation ( Gender 0.967 0.912 0.000 0.006 0.008 0.017
Relationships between gender and search place.In study place (%) In public (%) On the way (%) On vacation (%) Man 43.7 3.0 4.4 0 Woman 22.2 10.8 12.7 4.1
Personal attributes and social environment.Context Personal attributes ( Social environment ( Gender Age Grade Major Activity Relations Importance Portal Personal Gender - - - - 0.000 0.138 0.116 P attributes Age - - - - 0.357 0.231 0.573 0.717 Grade - - - - 0.399 0.786 0.396 0.202 Major - - - - 0.610 0.577 0.150 0.598 Social Activity 0.000 0.357 0.399 0.610 - 0.000 0.000 - environment Relations 0.138 0.231 0.786 0.577 0.000 - 0.633 0.171 Importance 0.116 0.573 0.396 0.150 0.000 0.633 - 0.573 Portal P 0.717 0.202 0.598 - 0.171 0.573 -
Relationships between gender and activity.Activity Top 1 (%) Top 2 (%) Top 3 (%) Gender Man On break (55.6) Studying (37.0) Others (4.4) Woman On break (61.9) Studying (18.4) Others (7.9)
Gender and search portal. < 0.05 < 0.05Search portal Gender ( Browser 0.001 Video software 0.239 Music software 0.627 Transportation software 0.036 Social software 0.000 Shopping software 0.021 Application store 0.624 Education software 0.798
Relationships between activity and relations.Alone (%) With others (%) On break 49.3 - Working 55.0 - Studying 50.0 - Shopping - 42.9 Traveling - 85.7 Others - 38.7
Relationships between activity and task importance.Important (%) Unimportant (%) Neutral (%) On break 15.9 38.1 45.9 Working 45.0 10.0 45.0 Studying 40.7 11.1 48.1 Shopping 28.6 0 71.4 Traveling 50.0 21.4 28.6 Others 29.0 38.7 32.3
Physical environment and social environment.Context Physical environment ( Soial environment ( Time Place Weather Device Activity Relations Importance Portal Physical Time - - 0.736 - 0.593 0.586 0.206 - environment Place - - 0.415 0.866 0.000 A - - Weather 0.736 0.415 - 0.593 0.559 0.374 0.412 - Device - 0.866 0.593 - 0.137 0.037 0.061 0.186 Social Activity 0.593 0.000** 0.559 0.137 - 0.000 0.000 - environment Relations 0.586 A 0.374 0.347 0.000 - 0.633 0.171 Importance 0.206 - 0.412 0.631 0.000 0.633 - 0.573 Portal - - - 0.186 - 0.171 0.573 -
Relationships between place and activity.Activity Top 1 (%) Top 2 (%) Top 3 (%) Place At home or dormitory On break (80.5) Studying (13.4) Working (3.7) In workplace Working (50.0) On break (31.3) Studying (12.5) In study place Studying (58.9) On break (34.1) Others (4.7) In public On break (44.7) Others (28.9) Shopping (10.5) On the way On break (50.0) Traveling (23.9) Others (23.9) On vacation Traveling (84.6) On break (15.4) -
Place and relations. < 0.05Place Relations ( At home or dormitory 0.001 In workplace 0.000 In study place 0.000 In public 0.037 On vacation 0.000
Relationships between place and relations.Top 1 (%) Top 2 (%) Top 3 (%) At home or dormitory Alone (50.4) With classmates (23.2) With friends (19.5) In workplace Alone (43.8) With friends (31.3) With workmates (12.5) In study place Alone (47.3) With classmates (38.0) With friends (10.9) In public With friends (42.1) Alone (26.3) With classmates (23.7) On the way Alone (60.9) With friends (32.6) With strangers (7.7) On vacation With friends (92.3) Alone (7.7) -
The personal attributes of the participants and the physical environment did not correlate with each other except for gender and place. We made further analysis of the two dimensions (Tables 24 and 25).
Tables 24 and 25 show that the four kinds of places were closely connected with gender. Specifically, men tended to do mobile search in study areas, whereas women easily performed mobile search when they were on the way to places, in public, and on vacation. Mobile search was more popular in women’s daily life, while men preferred to use mobile devices when studying.
The detailed correlation results are listed in Table 26.
The detailed correlation results are listed in Table 31.
The search place and the relations of the participant were significantly correlated. To make it as specific as possible, we performed across tabulation analysis between every place and the relations (Tables 33 and 34).
As shown in Table 33, there are significant correlations between search places and current relations when doing mobile search and Table 34 shows the correlations more specifically. The participants were alone when at home or in a dormitory, in their workplace, and in a study place. They were often with their friends when in public or on vacation. This has something to do with the characteristics and functions of the places.
Based on the analysis of the impact of mobile search motivations and context on search behaviors, we built a multi-dimensional model of mobile search behaviors and the factors that influence it (Figure 1).
The motivation- and context-based mobile search behavior model.Figure 1
Search motivations affect mobile search behaviors in terms of mobile search sessions, mobile search queries, and information needs. In contextual dimensions, the time and place of mobile search correlate with the mobile search query. In addition, gender, search activity, search task importance, and portal all have significant correlations with the information need types. The correlations among the contextual dimensions are also significant. Gender, time, place, search activity, relations, task importance, and portal correlate with each other in several ways, and the other dimensions have no significant correlations. In short, mobile search is a successive process involving mobile search motivation, context and mobile search behaviors.
From the listed findings in the mobile search session and information motivation, we note it is hard to differentiate curiosity-based search sessions from those based on recreation or learning, because users tended to find positive reasons for time killing and recreation. Another possibility is that it is difficult to distinguish among the three motivations in participants’ cognition because recreation and learning are at times incorporated or have blurred lines between activities and are now harder to recognize or categorize than before. Participants may need information to satisfy curiosity, to increase knowledge, or to enjoy information and recreation, and users rely on mobile phones to meet all these needs. However, life service emphasizes the instant usefulness of information, such as travel information and data to help users make comparisons and decisions on consumer and other activities.
When it comes to mobile search query, we find that queries coming from life service motivations are generally uniform and hard to change, indicating that information needed for life service is more conventional than mobile searches driven by curiosity, time killing, and learning.
The motivation for doing mobile search has significant correlations with the information need types, especially when driven by curiosity, learning, and enjoying life service. However, the correlation between emotion change and motivation is not strong in this study. As can be expected, under different search motivations, users appeared to need diverse information types. The need for culture and art is shown to be the highest. College students are commonly committed to their studies, however, and information searches are often motivated by curiosity and learning.
The mobile search context has significant correlations with mobile search behaviors, which can be explained from four aspects: the search session, the search query, the information need, and the emotion change after searching. There may be some reasons for these behaviors hidden behind the correlations. As use of the mobile Internet is becoming pervasive from early ages, college students have gained a good command of it, and thus use mobile search every day. Also, college students are familiar with the mobile Internet, and their daily life on campus reflects more diversity in terms of interests, lifestyles, and opportunities, and their schedules while often full will have flexibility. As college students also place high importance on social activities and allow for discretionary time in their schedules, they have strong and diverse information needs. In addition, we notice that browsers are found to be a better choice than other portals for mobile search. Search engines accessed through browsers are very general, and useful for all kinds of searches.
There are also significant correlations among the different mobile search context dimensions, such as gender with place, search activity, and portal. Men and women exhibited different daily routine priorities and information concerns, as women tended to stay at home more than men who often preferred to go outside. What’s more, women in this study were more engaged in social and consumer activities, while men’s interests tended towards the technical and scientific concerns. Furthermore, college students tended to do mobile search in certain situations, and often their searches are triggered by certain activities. For instance, they tended to search for information for learning when studying something new.
Our study adopts a free experiment to learn about college students’ mobile search behaviors in their daily life, where participants contributed information on their phone logs and diaries to help us understand how search motivations and context influence their mobile search behaviors. Conclusions follow.
About three-fourths of the mobile search sessions are motivated by a single motivation, while a quarter are motivated by multiple motivations. Different search motivations have the tendency to cross and converge, especially among searches driven by curiosity, time killing, and learning. Queries in search sessions influenced by the former three motivations tend to be repeated less frequently than queries in life service, which suggests that the former three motivations stimulate more types of information needs, while the life service motivation is more stable and uniform. Information needs are found to be mainly driven by three types of motivation: curiosity, learning, and life service. No significant correlations are found between emotion changes and motivations.
We find that information search behaviors significantly correlate with the multi-dimensional context and the motivation. Gender, search activity, task importance, and portal all have significant correlations with the information need type, and the information needs driven by different motivations are diverse. Furthermore, the correlations among the dimensional context are significant. Gender, time, place, search activity, relations, search task importance, and portal dimensions are all correlated in several ways, and the other dimensions have no significant correlations.
A limited number of participants and brief experimental duration are the main limitations in our research. Given a shoestring budget and the need for remuneration of every participant, the experiment lasted only half a month. In future studies, we will make comparison with studies of desktop search behaviors, and focus on the impact of subjective factors on mobile search behaviors. We may also look at sample groups other than college students and in other locations to make the study more applicable across various populations.