Meditating differences in children's interaction with digital libraries through modeling their tasks
Describes four graphical preliminary task-based models of ten Arabic-speaking children, ages 6-10, based on their interaction with the International Children's Digital Library (ICDL). Data generated from a previous study (Bilal & Bachir, 2007b) were coded and analyzed to generate the models. Seven modes of behavior characterized children's interaction: Start, Recognize, Browse, Differentiate, Read, Explore, and Finish. Each mode is associated with moves based on task characteristics. The models were constructed using the general model developed by Bilal, Sarangthem, & Bachir (2008), which was partially informed by the works of Ellis (1989), Ellis & Haugan (1997), Choo, Detlor, & Turnball (2000), and Marchionini (1995). New patterns of behavior that are missed in these works were identified (Explore and Read). The models lack a Search mode that characterized children's keyword searching. The ICDL allowed entry of Arabic script in the search box, but failed to retrieve Arabic books by keyword. Children's behavior that combined linear and non-linear progression and the core iterative processes that occurred between certain modes of behavior and varied by task provide additional perspectives for understanding information seeking behavior within the specific context of a small and well structured Web space such as the ICDL.
Many studies have investigated the relationships between a user task and information seeking behavior (Xie, 2007; Jarvelin & Ingwersen, 2004; Jarvelin & Wilson, 2003; Ingwersen & Jarvelin, 2005; Vakkari, 2003; Bystrom & Jarvelin, 1995; Belkin, Chang, Downs, Saracevic, & Zhao, 1990; Saracevic & Kantor, 1988; Belkin, 1980). Researchers have investigated various types of tasks including collaborative vs. individual (Hyldegärd, 2009), open vs. closed (Bilal, Sarangthem, & Bachir, 2008); transactional vs. informational (Terai, et al. 2008); complex/goal-oriented (Bartlett & Neugebauer, 2008); directed vs. general purpose (Thatcher, 2008); factual vs. interpretive vs. exploratory (Kim, 2007); known-item vs. subject-oriented (Kim & Allen, 2002); imposed vs. self-selected (Gross, 2004); assigned vs. semi-assigned vs. fully self-generated (Bilal, 2002; Bilal, 2001; Bilal, 2000). Studies have also examined task specificity, monitoring, and interruption (Bailey, Adamczyk, Chang, & Chilson, 2006); task manipulation (Vakkari, 2003); and task complexity (Bystrom & Jarvelin, 1995; Vakkari, 1999). Recently, Wildemuth & Hughes (2006) have raised the issue of the embedded tasks. Despite these prolific works about task influence on information behavior, there are insufficient empirical studies that resulted in task-based graphical models of children's information seeking behavior in digital environments. This paper describes four preliminary task-based models of Arabic speaking children's interaction with the international Children's Digital Library (ICDL) and identifies the core iterative processes and linear progression children performed on each task. The models provide additional understanding of children's information behavior in digital environments that a general model of such a behavior may not provide.
The overarching research question addressed in this paper was: What task-based empirical models can be generated based on Arabic-speaking children's interaction with the ICDL?
Due to space limitation, only two bodies of relevant literature are briefly reviewed: 1. Models of children's information seeking behavior, and 2. Children's use of the ICDL.
1. Models and children's information seeking
There is scarcity of empirical work that modeled children's information seeking behavior, in general, and in using Web-based interfaces (e.g., digital libraries), in particular. This section reviews studies that concluded with models of children's information seeking behavior in different Web environments.
Using data sets collected of a previous study that investigated Arabic-speaking children's use of the ICDL (Bilal & Bachir, 2007b), Bilal, Sarangthem, & Bachir (2008) developed a preliminary, general model of the children's behavior that consists of 7 modes of information behavior: Start, Recognize, Browse, Differentiate, Read, Explore, and Finish. The model is. partially informed by the work of Ellis (1989), Ellis & Haugan (1997), Choo, Detlor, & Turnball (2000), and Marchionini (1995). One limitation of this model resides in its lack of a Search mode. Children used Arabic script in the ICDL keyword search box, but the ICDL failed to retrieve books due to its lack keyword indexing of Arabic books. This deficient design feature of the ICDL provided an incomplete representation of the children's interaction.
Shenton (2007) developed a graphical model of information seeking failure of young users, ages 3-18. The model consists of five dimensions: Source Dimension, Knowledge Dimension, Skills Dimension, Social Dimension, and Psychological Dimension. Each dimension lists the issues that contributed to information seeking failure. Relationships among the dimensions are clearly described. However, the model does not show the “task” as a factor in any of the dimensions.
The imposed query framework developed by Gross (1997) pointed out the issues in queries assigned by teachers in elementary schools. Use of the framework was investigated in varied information environments including adult reference desk in public libraries where Gross & Saxton (2002) examined the effect of task type on user assessment and satisfaction. The framework has demonstrated utility in informing practice and guiding professional training in reference services. However, it focuses on the imposed query rather than on all types of queries.
2. Children and the ICDL
A handful of studies exist about children's use of the ICDL. In a two-part study, Bilal & Bachir (2007a-b) examined the information seeking behavior of ten Arabic-speaking children, ages 6-10, in using the ICDL to find Arabic books on four different tasks. The study took place at Bibliotheca Alexandrina in late December 2004. In part one, Bilal & Bachir (2007a), assessed the cross-cultural usability of the ICDL as an international Web interface. In part two, Bilal & Bachir (2007b) investigated the children's information seeking behavior, success, and affective reactions in using the ICDL. Children were given four different tasks to perform in the ICDL: one assigned and fact-based, one assigned known-title and fact-based, one semi-assigned and research-based, and another fully self-generated. Children's interaction was captured online and their affective reactions were elicited during exit interviews. Children's information seeking behavior was characterized by browsing using a single function; that is, looking under Arabic language from the drop-down menu to view the Arabic book collection. Although children were able to type Arabic keywords using an Arabic keyboard, the ICDL returned zero hits due to lack of indexing of the books metadata. The data sets generated from this study were used to develop the task-based models described in this paper.
Massey, Druin, & Weeks (2007) examined the affective reactions of twelve children from four different countries to reading and reviewing books in the ICDL. Children were selected from the United States, Honduras, New Zealand, and Germany. Children were given a book review form to use for answering five questions about each book they read. Findings revealed that children preferred books with happy endings. Language capabilities were a stronger factor in choosing books than culture or nationality as children did not choose books in languages they did not speak. When children were unable to read books in different languages, they relied on images and colors to understand the story and to express how it made them feel. Therefore, their interpretations of book images influenced their emotional classifications of the books. The findings provided insights into the role of emotions in children's book selection and reading. They also informed us that visual representations of textual information could help overcome language barriers in book selection and reading.
In an earlier study, Reuter & Druin (2004) investigated the searching and book selection behavior of ninety-six first- through fifth-grade children from the suburbs of Maryland. Age and gender influenced searching and book selection. Younger children preferred simple and more interactive interfaces; whereas older children favored more sophisticated interfaces. Issues in system designed were identified and recommendations were made for making the ICDL more supportive of younger children.
The literature review showed a scarcity in studies that modeled the information seeking behavior of children in digital environments. The four task-based models described in this paper are a first step towards filling this research gap.
The task-based modelswere developed based on data sets generated from the previous study (Bilal & Bachir, 2007b) where both quantitative and qualitative methods were employed to generate the data.
Ten Arabic-speaking children ages 6-10 participated in data collection. Children were recruited by staff at Bibliotheca Alexandrina's Children's Library (Bib. Alex.). Four children were male and six were female. Based on the data generated from the open individual interviews, one child was six-years old, three were seven-years old, two were eight, two were nine, one child was nine and a half years-old, and one was ten-years old. Children possessed adequate level of computer and Internet experience. They were from different social background who attended private and public schools. Only one child was familiar with the ICDL. All children had limited English language skills. Children read books on a regular basis and were regular users of Bib. Alex.
Children were asked to perform these four tasks: 1. How many books does the ICDL have in the Arabic language, 2. Find a book in the Arabic language named Dima and open the first page of the book, 3. Find a book about animals in the Arabic language and write the name of the book on your sheet, and 4. Find any book in the Arabic language and read as many pages as you can.
The four task-based models were generated using the data sets generated from the previous study by (Bilal & Bachir (2007b). Children's interaction with the ICDL was captured online using HyperCam software package. Their affective states were gathered during an exit interview. Each child's ICDL interaction activity on each of the four tasks was saved, replayed, coded, and analyzed. Each child's ID number and interaction activity was input into an excel sheet along with the sequence of the activity, a description of the activity (e.g., scan, browse, select book, open page, move to next page), and comments and/or observations about the activity. This process resulted in a total of forty excel sheets on the four tasks (10 children x 4 tasks). These sheets formed the basis of the data sets that were utilized to build the models.
Models used as frameworks
The four task-based models were informed by the general model that represented the children's interaction with the ICDL (Bilal, Sarangthem, & Bachir, 2008). This general model was partially based on the work by Ellis (1989), Ellis & Haugan (1997), Choo, Detlor, & Turnball (2000), and Marchionini (1995). Certain modes and moves of behavior shown in the models were adapted from each of these authors’ work. New modes of behavior and associated moves that are missed in these frameworks were generated.
The models consist of 7 modes of information seeking behavior and moves associated with each of them. Each mode and move is labelled for ease of understanding of the models.
Start (M1): This mode characterizes the beginning of a task after a child recognizes the information need. It generally begins in the ICDL default interface by scanning the Simple search, Advanced search, and Keyword search features. The scan move (M1V1) results in making a selection that takes the child to the interface containing the book language options. The select move (M1V2) is confined to the Simple search interface. A child returns to the ICDL homepage to restart a task or to explore the other features.
Recognize (M2): This mode embraces two moves, scan and select. The scan move (M2V1) at this step consists of activating the book language drop-down menu from the ICDL Simple Search interface. The select move (M2V2) pertains to choosing the Arabic language option. This activity results in a display of Arabic books thumbnails.
Browse (M3): This mode comprises browsing and is of two types: Directed and Semi-directed. Directed browsing includes two specific moves, view (M3V1) and verify (M3V2). These moves typically characterize a behavior on a target-oriented task (e.g., finding an answer for a fact-based task). It also extends to browsing to find a known-title item (e.g., finding a book with a specific title in the Arabic book collection). Semi-directed browsing includes one specific move, examine (M3V3), that reflects a behavior on a less definite or less focused task (e.g., finding a book of interest for a semi assigned task or for a fully self-generated task). The examine move may be followed by filtering sources and making a selection decision.
Differentiate (M4): This mode consists of filtering books based on differences or criteria such as book cover, images, book length, subject matter, affect (happy/sad), or other factors. Differentiation is both Directed and Undirected. Directed Differentiation is similar to Directed browsing in that it is focused and target-oriented. It consists of the more specific view move (M4V1). Undirected Differentiation has very little focus and is described by the sweep move (M4V2), which reflects a behavior such as paging through book thumbnails, book titles, book pages, and/or jumping from one thumbnail to another. Typically, sweeping includes little reading or engagement in a task.
Read (M5): This mode involves different levels of reading and consists of two types: Directed and Undirected. Directed Reading is characterized by the view move (M5V1) that takes account of reading engagement. Undirected Reading includes little or no reading and has very little focus (flip). The flip move (M5V2) also reflects a much faster behavior than the sweep move.
Explore (M6): This mode embraces use of the navigation controls of the ICDL and/or Internet Explorer. It has two specific moves: Navigate (M6V1), which includes activation of all navigation controls, except for the Back button (Internet Explorer) and the various back icons (ICDL). All other navigation activities are classified under the backtrack move (M6V2).
Finish (M7): This mode marks a completion of a task. It sometimes results in note taking, as required by a task.
The seven modes that characterized children's information seeking behavior on the four tasks are seen in Figures 1–4. Children's information seeking on tasks 2-4 (Figures 2–4) shows 7 modes of behavior (Start, Recognize, Browse, Differentiate, Read, Explore, and Finish). However, the behavior on Task 1 (Figure 1) precluded the Differentiae mode due to the nature of the task that did not necessitate book filtering or making book selection decisions.
Children's information seeking behavior was dominantly non-linear and iterative and included core iterative processes between modes of behavior that varied by task. The exception was the behavior on the fact-based task (Task 1) that was mostly linear in nature. On this task, children were asked to locate the total number of Arabic books, which pops up as soon as the Arabic language option is selected from the Language drop-down menu that resides in the ICDL Simple search interface.
As seen in Figure 1, children's iterative transitions were minimal and occurred only between the Information Need that was driven by the task and the Explore mode (Information need-Explore-Information need). However, from Start to Finish, children's behavior on this task was typically linear between the modes and without iterations (Start-Recognize-Browse-Read-Finish). Children made the lowest number of moves (ToMV=64) and the lowest level of iterative transitions (ToT=30) on this task (Figure 1).
On Task 2, children's information seeking consisted of the seven modes of behavior and showed a higher level of interaction than on the previous task (Figure 2). Children's iterative transitions progressed between Browse-Differentiate-Browse; Differentiate-Read-Differentiate-Explore; Read-Browse-Read-Explore. The core iterative processes occurred b between Browse and Differentiate (T=13) and between Browse and Explore (T-12) and accounted for half of the total number of transitions made (ToT=50). The total number of moves children exhibited on Task 2 were higher than those on Task 1 (ToMV=78 vs. ToMV=64, respectively). Similarly, the number of specific Explore moves were much higher on this task, as opposed to the previous one (MV=13 and MV=3, respectively).
Children's behavior was much more interactive on Task 3 (Figure 3) than on the two previous tasks, as evident in the level of iterative transitions they made between these modes: Browse-Differentiate-Browse; Browse-Explore-Browse; Differentiate-Read-Differentiate; Differentiate-Explore-Differentiate-Start; Read-Explore-Read-Browse-Read. Here too, the core iterative processes occurred between Browse and Differentiate. Children made 18 transitions (T=18) from Browse to Differentiate and 11 transitions (T=11) from Differentiate to Browse, totalling 29 (T=29), slightly over one-third of the total number of transitions made on this task ((ToT=80).
The moves children made on Task 3 were much higher than those on Tasks 1 and 2 (ToMV=145 vs. ToMV=64 vs. ToMV=78, respectively). Similarly, the specific Explore moves they made were also much higher on this task, as opposed to Tasks 1 and 2 (MV=21 vs. MV=3 vs. MV=13, respectively). Children's reading reflected in the Read mode increased dramatically on Task 3 (MV=47) over Task 2 (MV=7).
Figure 4 reflects children's information seeking on Task 4. Children made nearly the same number of transitions on Task 4 as they did on Task 3 (ToT=79 vs. ToT=80, respectively). However, the total number of moves on Task 4 were much higher than those on Task 3 (ToMV=286 vs. ToT=145, respectively), due to the amount of reading (MV=187 out of ToMV=286) that was driven by the task demand.
Children's behavior on Task 4 flowed between Recognize-Differentiate; Browse-Explore-Browse; Browse-Read; Read-Recognize; Differentiate-Browse-Differentiate; Read-Explore-Read; and Explore-Differentiate-Explore. Again on this task, children's core iterative processes progressed from Browse to Differentiate (T=10) and from Differentiate to Browse (T=12). However, these processes were slightly lower (T=22) on this task than they were on Task 3 (T=29) and Task 2 (T=25).
New patterns of behavior
Children's information seeking exhibited new patterns of behavior that are missed in models used as frameworks. These include the Explore and Read modes. While the Explore mode is Web-specific in that it an integral component of navigating web space (Back, Forward, Home, etc.), the Read mode is task-specific. The amount of reading was mainly driven by each task characteristics and was influenced by the level of children's engagement with the content of the book they selected or encountered. Affective reactions that underlined the children's information seeking behavior are also absent in these frameworks.
Arabic-speaking children's task-based information seeking in the ICDL was characterized by seven modes of behavior: Start, Recognize, Browse, Differentiate, Explore, Read, and Finish. However, on Task 1, children's behavior precluded the Differentiate mode due to the fact that the task was closed and had a target answer (numeric information) that did not necessitate filtering books and making a selection decision. The simplicity of this task combined with the convenient ICDL display of the target answer (26) above the Arabic book thumbnails made easy to find by older children. Two younger children (ages 6-7) who did not pay attention to this information pursued a different approach- counted each Arabic book thumbnails and calculated the total number of books (used their fingers to count). This alternative strategy, although creative, resulted in miscounting the books (37 instead of 26) by one child.
The dominant linear behavior children exhibited on Task 1 (Figure 1) was not surprising, considering the small space allocated for the small Arabic book collection, sequential display of thumbnails, and immediate availability of the target answer. One should expect this type of behavior in using a small and structured Web space such as the ICDL. Conversely, in using a Web search engine to find information for a similar task, one should expect a more iterative behavior from children due to the heterogeneous nature of the Web (Bilal, 2000).
Children's behavior on the known-item title (Dima) (Task 2) was somewhat iterative. As seen in Figure 2, children's core processes flowed from Browse to Explore and from Browse to Differentiate. The centrality of these processes suggests that children continued to explore the collection even after locating the target book. This behavior can be explained by the fact that on Task 1, children did not have the opportunity to familiarize themselves with the content of the Arabic book collection, or that children were uncertain as to whether the collection had more than one book with the same title, Dima. It also insinuates that children began to construct knowledge about the content of the Arabic book collection and the ICDL various features as they began working on Task 2. The uncertainty children experienced in using ICDL was due to their perceived inability to use it successfully (based on data generated from exit interviews). This affective state is typical of user behavior and emerges due to lack of familiarity with the ICDL. Kuhlthau (2003) supports this view by noting that “when the information search process is viewed as a process of construction, uncertainty and anxiety are anticipated and expected as part of the process” (p.7).
The behavior children exhibited on Task 3 (Figure 3) was more iterative than that seen on Task 2 (Figure 2). Previous studies found that the more open-ended a task was, the more browsing moves children made (Large & Beheshti, 2000; Schacter, Chung, & Dorr, 1998). This was, indeed, the case in this study as children's browsing moves (shown under the Browse mode) on Task 4 were slightly higher (MV=35) than those on Task 3 (MV=31). The number of iterative transitions children made on Task 4 was nearly the same as that on Task 4. This finding was not expected as Task 4 was more open-ended, fully self-generated, and most complex. Finding a book of interest on the semi assigned topic (Animals) for Task 3 seemed to be slightly more complex for the children than Task 4. To select a book about the topic, children had to: a. identify the books in the Arabic collection dealing with the assigned topic, b. employ filtering skills, and c. make a book selection decision based on interest. The fact that children's core iterative processes occurred between Browse-Differentiate and Differentiate-Browse and the fact that the number of moves they made (29) was the highest across all tasks substantiate the higher complexity level of executing this task.
Figure 4 represents children's information seeking behavior on the open-ended, fully self-generated task (Task 4). Children were asked to find a book of interest on any topic and read as many pages as they could. As avid readers, two children read two books instead of one and the rest completed one book each. The amount of reading children exhibited on this task was attributed to task demand that, in a few cases, children exceeded its requirements. Similar to the reading behavior on Task 3, children ‘s reading on Task 4 was both Directed (View) and Undirected (Flip).
The level of iterative transitions children made on Task 4 was nearly the same as that on Task 3. Similarly, the number of moves associated with each of these modes was nearly the same as those shown on Task 3, with the exception of the amount of reading that was much higher on Task 4 (Figure 4). The iterative core processes that occurred between Browse- Differentiate and Differentiate-Browse were slightly lower on Task 4 than those on Task 3. By the time children performed Task 4, they had already been exposed to the Arabic book collection, constructed some level of knowledge about its content, and gained experience in manoeuvring within the ICDL environment. These skills and knowledge gain could have contributed to the children's effectiveness in performing Task 4.
The task-based models identified depict the combined nature of the children's linear and non-linear iterative processes. Absence of one of the seven modes of behavior (Differentiate) on the fact-based task testifies to the fluidity of the stages of the information seeking process and, thus, provides additional perspectives about the influence of task characteristics on the process of information seeking.
The limitations of this study may reside in using a small (10) and convenient sample. Like any research involving human subjects, obtaining a random sample was difficult to achieve. Therefore, the information seeking behavior reflected in the task-based models may not be representative of all Arabic-speaking children. Lack of a Search mode in these models was due to the design of the ICDL that, although it accepted Arabic script in the keyword searching box, it failed to retrieve Arabic books. Testing the validity of the models using the Arabic version of the ICDL could provide data about keyword searching that would contribute to building this important component. In addition, allowing children to use the entire book collection rather than limiting them to the Arabic collection could provide a more holistic picture of their interaction, book selection, and book reading strategies.
This study aimed at identifying task-based empirical models that could be generated based on data sets collected in a previous study of Arabic-speaking children's information seeking behavior in the ICDL (Bilal & Bachir, 2007b). The data on which these models are based were collected in late 2004 when the ICDL interfaces could only be negotiated in the English language. Currently, most of the interface design of the ICDL has changed and versions in many languages, including Arabic are available. Validating these models in future research should account for language and changes in the interface design.
The task-based models are a crossover from the grounded works of Ellis (1989), Ellis & Haugan, (1997), Marchionini (1995), and Choo, Detlor, & Turnball (2000), yet they unveil new patterns of interaction (Explore and Read) that should be explored further in future research. Children's information seeking was characterized by seven modes of behavior: Start, Recognize, Browse, Differentiate, Read, Explore, and Finish. Each mode consisted of moves that varied by task. On the fact-based task, however, children's behavior precluded the Differentiate mode as it was strongly influenced by task characteristics. This finding affirms the fluidity of the nature of the information seeking process and that these modes of behavior are not fixed in space or time.
Children's behavior that combined linear and non-linear progression and the core iterative processes that occurred between certain modes of behavior and varied by task provide additional perspectives for understanding information seeking behavior within the specific context of a small and well structured Web space such as the ICDL. Further researchis needed to validate the models and to unveil the nature of the children's searching strategies. Using a large sample of children from various geographic locations, allowing them to use the whole ICDL book collection, and constructing tasks with various layers of complexity could provide a more valid representation of children's information seeking behavior in the ICDL.
This paper is based on a research project that was partially funded by the University of Tennessee Office of Research. The authors wish to thank the University for its financial support. The comments and suggestions of thereviewers are greatly appreciated.