A participatory design approach to information architecture design for children

Abstract

A design process that involves the participation of children should effectively elicit the needs of child users by considering characteristics such as their age-appropriate levels of cognitive ability, shyness, language skills, and motor-sensory capabilities. Since the 1960s, a variety of methods and guidelines have emerged specialising in children. Participatory design deals with the problem of enabling users to participate in the design process and with the task of generating ideas by means of generative toolkits and workshops. Hence, participatory design enables designers to look at problems from a child's standpoint and, simultaneously, deal with childhood traits such as shyness and immature language skills. This paper introduces two participatory design toolkits: namely Info Block and Info Tree. These toolkits enable users to build information architecture (IA) that can, firstly, reflect their cognitive characteristics and, secondly, elicit user needs with respect to the information architecture design of children's websites. In a case study, the toolkits were used to evaluate the usability of the Yahoo!® Kids (Korea) directory. The results show that the information architecture of children differs from that of adults in depth, breadth, and clarity of contents and logicality. Finally, suggestions are made to improve the usability of children's websites.

Keywords:

• children
• information architecture
• participatory design

1. Introduction

In information architecture design, conventional methods of studying the participation of users, such as focus group interviews, usability testing, and log data analysis, are less effective for determining the needs of children because the methods, which were developed primarily for adults, ignore the developmental characteristics of children, such as shyness and limited language and social skills. Of the many design methods that have been developed for children, participatory design is useful for eliciting the tacit and latent needs of users, and can thereby provide abundant data on the cognitive characteristics that need to be supported in the design of information architecture.

Rather than relying on the individual creativity of designers to solve design problems in a specific context, a participatory design process relies on the collective generativity of stakeholders; in other words, it uses the collective ability of stakeholders to generate or create thoughts and imaginings. By analysing what users create with generative toolkits, researchers can elicit the tacit needs of users; that is, the needs that are implied or indicated but not actually expressed.

In March 2000, Yahoo! Korea launched a portal website for children, called Yahoo!® Kids, Korea (in short, Yahoo Kids). Yahoo Kids, which has been one of the leading websites for children, with over 15 million page views per day in 2002, is a good case study for testing the participatory design approach because it targets children and has not previously been exposed to the participatory design approach. Yahoo! Korea listens to customers through various channels, such as interviews with parents and teachers and feedback from website users; in most cases, however, the feedback is from adults, not children. In 2002, the company conducted an in-lab usability test but the results showed that children were rather passive and bored throughout the experiment. Like Yahoo!, Yahoo Kids provides a web directory or categorisation of links to other websites. Yahoo Kids directory has numerous links of websites, which are displayed by category and subcategory, making it a good example of hierarchical information architecture.

Because our case study is concerned with the usability of a website, we had to develop toolkits that help children express their cognitive processes in a form that can be analysed. Hence, we developed two types of generative toolkits, Info Block and Info Tree, both of which help children create information architecture for themselves. The toolkits are based on the psychological observation that play is a child's natural medium of self-expression (Axline 1969) and that children are most creative when playing. Thus, the toolkits resemble children's everyday toys in form and function, and were designed to stimulate natural behaviour during tests. While the two toolkits have a common goal of improving the user interface of Yahoo Kids, each serves a different purpose. Info Block focuses on analysing the existing information architecture, whereas Info Tree focuses on discovering the interests and expectations of children. By analysing the information architecture that children built with the toolkits, we can identify its characteristics and compare it with the current information architecture of the adult-designed Yahoo Kids directory.

In section 2 of this paper, we explain the background to our research and discuss why information architecture design for children is an important issue in the market. In section 3, we describe conventional methods of studying users and discuss the problems that arise when these methods are applied to children. In section 4, we introduce the participatory design approach and compare it with conventional methods. In section 5, we describe a case study and explain the development of the toolkits, as well as the experimental process and results. In section 6, we discuss how information architecture of children differs from that of adults, and we explore the participatory design process of the experiment. Finally, we present our conclusions and the limitations of the study in sections 7 and 8.

2 Emergence of children as a major user group

The emergence of the Internet and a wide range of software products for children have exposed users to complex forms of information architecture. In many developed and developing countries, where Internet access is available at home, at school and at Internet cafes, elementary school students have emerged as enthusiastic users. This phenomenon has led to a boom in e-businesses that targets children, including online communities, games, search engines and educational services. Online educational services, for example, are merging with the educational software industry, and online games and virtual communities are popular among children, even causing a social problem of PC addiction (Baek 2003). In the case of South Korean search engines, there are at least six portal websites just for children, and all the sites provide links to numerous online educational services, games and communities.

McNeal (1992) presents the following reasons for the growth of the children's market: first, the children's market is the primary market where children can purchase things for themselves; products such as gaming software belong to this category. Second, the children's market is one where children can be stimulated to influence their parents' purchasing behaviour: a child's preference for educational software, for example, is likely to have a strong influence on which product the parents purchase. Third, the children's market is the future market where children will become major consumers in 10–20 years. Any interface or brand that children are familiar with will have an advantage over competitors in the future. Due to the increasing value of the children's market, we need to accurately and objectively elicit children's needs. Hence, marketing researchers have introduced numerous ways of evaluating the consumer behaviour of children (for example, observation, laboratory experiments, and role-playing) as well as ways of understanding the child's mind (for example, attitude scales, focus group interviews, and picture-drawing techniques; McNeal 1992). In the study of human-centred interaction, researchers began involving children in the design process as early as the 1960s; since then, the role of children has evolved along with the methods of eliciting their needs (Druin 2002). As the buying power of children continues to grow, so too does the demand for effective methods of studying users.

Despite the potential of the children's market, the e-business industry is reluctant to explore the type of user-centredness and user-participation that is pervasive in other fields of design. In cooperation with Yahoo Kids, our focus is on the participatory design of information architecture for children.

3 Conventional methods of information architecture design

For information architecture design, conventional methods of studying users include focus group interviews, usability testing, online questionnaires, user analysis, log files and statistic reports. The data collected from such methods are analysed for the purpose of finding problems and improving products. However, these methods are problematic when applied to children. First, children differ from adults in terms of developmental characteristics, particularly with respect to short-term memory span, language skills, and cognitive processes (Yoo 2000); hence, conventional methods are likely to present problems in effectively eliciting children's needs. For example, a child's shyness or passive attitude to unfamiliar people or an unfamiliar environment can obstruct communication with the tester. Second, because children participate in the design process at the evaluation phase, their influence on the decision-making process is limited. Druin (2002) claims that the later children participate in the design process, the less direct impact they have on changes in technology; moreover, when given a limited schedule, adults tend to decide what should be done in the end. On the other hand, when children participate in the earlier stages of the design process, there is more flexibility regarding when and where activities take place and the design of technologies is therefore likely to be more user-centred. Third, because of inadequate language skill, children may not express their thoughts clearly; hence, the task of discerning their needs from conversation is impoverished, especially when the conversation is about an abstract topic such as information architecture. In this case, analysis of user behavioural patterns or prototypes can supplement language-dependent methods and help discern the user's tacit needs (Sanders 1999; Druin 2002). As discussed in the following section, participatory design overcomes the limitation of verbal protocols by analysing what users have made.

4 Participatory design

While traditional methods of researching design focus primarily on observational research and questionnaires, participatory design focuses on the things people make in order to elicit what they think, feel and dream (Sanders 1999).

Figure 1. The relationship between action and need (Sanders 1999).

To solve design problems, participatory design uses the collective generativity1 of stakeholders rather than the individual creativity of designers (Sanders 1999). Designers can discover problems that are specific to a particular context by having workshops and discussing design issues with users, managers, and sales people. In this kind of workshop, designers use a generative toolkit to connect the ideas of people from different disciplines and perspectives. Participants of the workshops (the stakeholders) use toolkits as a form of ‘quick-and-dirty’ prototyping, to visualise their thoughts in the generative phase of a design process; designers then analyse the results to elicit the needs of the participants (Sung et al. 2003). Figure 1 illustrates the relationship between actions and the needs that may be extracted from them (Sanders 1999). Whereas the user needs collected from conventional methods are based on explicit knowledge or observable behaviour, the needs elicited from participatory design are based on tacit knowledge, which cannot readily be expressed in words (Sanders 1999). In Figure 2, a group of housewives are using a generative toolkit to modify a pram at a shopping mall. Participatory design is especially useful in cases like this where the design depends heavily on the context of the user (a mother carrying an infant), as well as the situation (shopping) and the environment (a shopping mall).

Figure 2. A group of housewives participating in a workshop for a baby carriage.

Participatory design with children was explored by Druin (1998). Druin claimed that when a participatory design method is applied to children it is possible to identify new technological possibilities that might not otherwise have been considered. Druin emphasised the importance of collaboration between children and adults and insisted that children who are not well-skilled in the development process could be inspired and empowered by their collaboration with adults to generate new ideas. Because low-tech tools give equal footing to adults and children as design partners, and because adults and children both know how to use these prototyping tools, children can actively participate in the process. Table 1 compares the participatory design approach and conventional approaches.

Table 1. Comparison of the participatory design approach and conventional approaches.

	Conventional approaches	Participatory design approach
Children's role in the design process	Children participate as a tester, i.e. they do the tasks given by researchers and they provide data	Children participate as a design partner, i.e. they participate in the design process and create their own information architecture
Dependence on children's language skills	Conventional approaches, such as focus group interviews, questionnaires, and in-depth interviews, depend heavily on language skills to elicit needs	Participatory design targets the eliciting of a user's tacit needs, relying to a much lesser degree on conversation
Children's behaviour pattern	Tasks are conducted alone in a laboratory environment with unfamiliar people; such situations can inhibit the active participation and natural behaviour of children	Since children do the tasks in groups and in a familiar situation, they tend to overcome the inhibition of the testing situation and behave more naturally
Dealing with children's cognitive process	Current approaches focus on the discovery of design problems by using usability testing and log data analysis; such methods are problematic because they make no allowance for the cognitive abilities of children	The information architecture of children can be compared with that of adults, leading to the discovery of fundamental differences between the two.
Educational effect	There is no significant educational effect	Teamwork and the process of organising information contribute to the development of children's social skills and the acquisition of knowledge
Element of fun	There is no element of fun	The design of the toolkits is based on the developmental characteristics of children and awareness of the value children attach to the pleasure of play.

Table 2. Composition of the teams.

Team	Toolkit	Grade	Age	Gender	Size
1	Info Tree	3	9	F, M	2
2	Info Tree	4	10	M	3
3	Info Tree	5	11	M	3
4	Info Tree	5,6	11–12	F	3
5	Info Block	3	9	F, M	3
6	Info Block	4	10	F, M	4
7	Info Block	5	11	F	3
8	Info Block	5,6	11–12	M	3
9	Card sorting	5	11	M	2

In this study, we used the participatory design approach to assess the usability of Yahoo Kids directory. We chose participatory design for three reasons. First, information architecture is too complex and abstract to discuss with children. Children have immature language and social skills, and may have difficulty articulating their needs when engaged in highly cognitive activities. However, they are accustomed to visualising ideas by drawing or making things, as these methods of generating ideas are typical childhood pastimes. This experience is why participatory design stimulates interest among children and helps them overcome passive attitudes or shyness when in the company of authority figures, a phenomenon frequently observed in the testing of children. Second, because cognitive activities are more difficult to observe than physical activities, cognitive needs are harder to elicit than physical needs. In the participatory design process, however, children use generative toolkits to produce prototypes that contain abundant data on cognitive characteristics. Third, because participatory design approaches treat design problems from the child's standpoint, they tend to overcome adult attitudes towards children. With its element of fun, the participatory design approach has several advantages, particularly with respect to the natural behaviour patterns of the children and their early participation in the design process; moreover, this approach is less dependent on language skills and more able to deal with variations in cognitive development.

5 Experiment

The purpose of this experiment was to collect information architecture created by children and to compare it with that of Yahoo Kids Directory so as to identify and analyse the cognitive characteristics of children. Before the experiment, we conducted a pilot test to validate the ability of children to comprehend the notion of information architecture and we gave them toolkits to create the information architecture.

5.1 Pilot test

In the pilot test, we asked 5 children (three from grade 3 of age 9 and two from grade 4 of age 10) to complete a set of tasks in 45 min. We gave each of them 35 keywords and asked them to organise the words in the form of a concept map. Next, we divided them into two groups and gave them toolkits (Info Tree and Info Block) along with instructions on how to use them. We then asked each group to design information architecture for a portal website, similar to Yahoo Kids. The results were analysed in terms of logicality and consistency of grouping. Four children successfully designed hierarchical information architecture in a manner that respected the natural grouping of the keywords. However, a girl from grade three merely arranged the keywords linearly and failed to form any hierarchical information architecture. In the group task, both groups created hierarchical information architecture with the toolkits. The test ended with a brief interview with the children. They said they had learned and practiced mind mapping in grade two, which probably contributed to their successful performance.

Figure 3. Hierarchical information architecture drawn by a third-grade girl.

Figure 4. Linear information architecture drawn by a third-grade girl.

5.2 Experimental method and procedure

The test involved 26 children (14 boys and 12 girls) aged from 10 to 12 years of age. We divided them into nine teams of two to four members and allocated an adult moderator to each team. Four teams were given the Info Tree toolkit and entrusted with the task of building information architecture for a children's portal website. They started by generating the major categories of the web directory and then added sub-categories to form hierarchical information architecture. Another four teams were given the Info Block toolkit and asked to reorganise the current information architecture of Yahoo Kids. Unlike the Info Tree teams, the Info Block teams were given the labels of the links provided by Yahoo Kids. On the Yahoo Kids website, the links lead to other websites and constitute the lowest hierarchical level of the information architecture. The task of the Info Block teams was to organise the links to reproduce the information architecture of Yahoo Kids. The ninth team was given the same task as the Info Block teams but instead of the Info Block toolkit, they used the ordinary card-sorting method. However, the result of the ninth team is not discussed in this paper because we think it is beyond the boundary of the focus of this study.

The adult moderator in each team collaborated with the children throughout the experiment. The role of a moderator was to help the children create ideas and encourage them to build their information architecture. The nine moderators, aged from 21 to 23 years of age, were industrial design graduate students. Their role is elaborated in section 5.3.2.

The experiment was conducted in a classroom rather than a lab to minimise the effect of the test conditions on the behaviour of the children. The test lasted 80 min and, with the children's consent, the process was recorded on video.

Figure 5. Experiment procedure.

5.3 Toolkits

The participatory design toolkits Info Block® and Info Tree® 2 analyse cognitive processes in order to identify user needs. The toolkits were intended to help the children build an information architecture in an enjoyable and easy manner (Axline 1969). In addition, they were designed to enable children to produce prototypes that reflect their tacit needs, thereby disclosing information about their cognitive processes (Sanders 1999). The toolkits also minimise problems associated with the passivity and shyness of children and a heavy dependence on conversation.

In designing the toolkits, we adapted the principle of play therapy. According to Axline (1969), play therapy is based on the fact that play is a child's natural medium of self-expression; just as adults ‘talk out’ their difficulties in therapy, children ‘play out’ their feelings and problems. Furthermore, Bruner (1972, cited in Yoo 2000) claimed that play contributes to the development of a child's creativity and flexibility. In play, the means are more important than the end; hence, in play a child is free from the burden of achieving goals and can consequently be more adventurous and experimental. Accordingly, if we apply the principle of play to the process of design for children, we can help the children actively participate in experiments and express their tacit needs. Moreover, because children in the concrete operational stage of cognitive development can play games with rules (Yoo 2000), we designed Info Tree and Info Block to resemble construction play 3 with rules.

5.3.1 Info Block

Info Block enables children to generate ideas and to use these ideas to construct hierarchical information architecture. The tool consists of 40 mm × 40 mm × 20 mm blocks of 7 spectral colours, each of which represents a depth of information hierarchy. Velcro is attached to the surface of the blocks so that the blocks can be easily grouped and stacked.

Figure 6. An Info Block and its stack.

The instructions on the use of the Info Block toolkit are as follows.

1. Form a group of three or four members and sit around a table.

2. Randomly put the blocks on the table.

3. Generate ideas related to a given subject and write them down on the red blocks.

4. Group the related blocks.

5. When the grouping is finished, generate keywords that represent the groups and write them down on the orange blocks.

6. Stack the grouped blocks and place the keyword block on top of each stack. An orange block should be placed on each stack of red blocks.

7. If the stacks can be grouped further, group the related stacks and generate keywords that represent each group. Write them down on the yellow block and put it on top of each group of stacks.

8. Repeat step 7 until no more groups can be formed. The colour of the next keyword block will be green, then blue and so on in the order of the seven spectral colours.

The instructions were originally given verbally during the orientation and delivered in Korean. In the experiment, we tried to use terminologies appropriate to the children. For example, instead of using the word ‘information architecture’, we used ‘mind map’ because children were familiar with the latter from the school curriculum. Neither was the term ‘hierarchy’ used. Instead, we illustrated an example of hierarchical information architecture on the board.

Figure 7. An example of prototype created by children with Info Blocks.

Figure 8. An example of information architecture created by children with Info Blocks.

5.3.2 Info Tree

Info Tree is based on Buzan and Buzan's principle of mind maps (Buzan and Buzan 1996) and provides children with a quick and easy way to generate and organise ideas. Info Tree uses the metaphor of a tree to help children comprehend the concept of information architecture; the tree consists of a Styrofoam trunk, wooden sticks and Styrofoam balls of different sizes.

Figure 9. The trunk and branches of Info Tree.

The instructions on the use of the Info Tree toolkit are as follows.

1. Form a group of three or four members and sit around a table.

2. Place the trunk on the table. The trunk symbolises the main topic of the information architecture.

3. Generate ideas that are related to the topic and, using post-it notes, write the ideas on the biggest Styrofoam balls.

4. Generate sub-ideas that are related to the ideas in step 3 and, using Post-It notes, write the sub-ideas on the next biggest balls.

5. Repeat step 3 until the Info Tree is complete.

Figure 10. An example of prototype created by children with Info Tree.

Figure 11. An example of information architecture created by children with Info Tree.

5.4 Results

The results of the sessions were analysed and evaluated in terms of the information architecture produced by the children and the participatory design process. For the information architecture, we quantitatively analysed the children's outcomes and made a comparison with the structure in Yahoo Kids directory. For the participatory design process, we made a qualitative analysis in terms of the following questions: Firstly, do the toolkits enable users to produce prototypes that reflect the tacit needs of users; that is, do they disclose information about the users' cognitive processes? Secondly, do the toolkits enable users to participate actively and unselfconsciously while at the same time reducing the heavy dependence on conversation and creating fun in the process?

5.4.1 Analysis of the information architecture

To identify the cognitive characteristics of the children, we analysed their information architecture in terms of logicality and structural shape. The logicality was assessed on the basis of every parent–child relation among the keywords in the information hierarchy. The structural shape was assessed on the basis of the breadth and depth of the information architecture. We applied the two criteria to the information architecture produced by the children and to the information architecture of Yahoo Kids.

Our analysis reveals the following characteristics of the children's information architecture.

	It has more illogical errors than the Yahoo Kids information architecture.
	The contents are categorised ambiguously; hence, a piece of information may belong to more than one category.
	It is wider and shallower than the Yahoo Kids information architecture.

With regard to the first characteristic, namely that the children's information architecture has more illogical errors than the Yahoo Kids information architecture, we categorised the following five types of parent–child relations.

	Inclusive: Keywords are in an inclusive relation if a parent keyword includes a child keyword by definition. For example, the relation between ‘The seasons’ and ‘Summer’ is inclusive because summer is one of the four seasons, but not vice versa.
	Illogical jump: There is an illogical jump in a relation if a parent keyword is too broad in meaning to include a child keyword. For example, the relation between ‘Sports’ and ‘Tiger Woods’ belongs to this category because other intermediate keywords can be used, such as ‘Athlete’.
	Inversely inclusive: Keywords are in an inversely inclusive relation if a child keyword includes a parent keyword. For example, the relation between ‘Summer’ and ‘The seasons’ is inclusive because the notion of seasons includes the notion of Summer, but not vice versa.
	Identical: Keywords are identical if a parent keyword and a child keyword have the same meaning. An example would be the two keywords ‘Education’ and ‘Education’.
	Associated: Keywords are associated with each other if their relation does not belong to any of the four categories above. The words ‘The seasons’ and ‘Travel’ would be an example of this category.

Of these relations, only the inclusive relation is considered a logical parent–child relation in hierarchical information architecture. An associated relation occurs as a result of the children's lack of logicality and their misunderstanding of conventional hierarchies. During the experiment, the children often grouped associated concepts and thought the organisation was complete. Illogical jumps are the result of the children's imperfect knowledge. For example, children may put a number of movie titles under the movie category because either they do not know that a movie can be further categorised into genres or they simply forget to complete the grouping. Inversely inclusive relations and identical relations occur mainly because children do not understand the meanings of parent or child keywords.

Table 3 shows that 50% of the children's structure has errors whereas only 11% of the structure of the current Yahoo Kids directory has errors. In the case of the current Yahoo Kids directory, the errors were intentional: that is, some popular keywords were used intentionally to enhance the search. For example, the ‘Game’ category in the main directory is linked to a page that lists the genres of games and the six most popular games. In this case, the six most popular games are emphasised. The children's structure has twice as many illogical jumps as the current Yahoo Kids directory. This phenomenon shows that, with respect to creating information architecture or navigating websites, the children's thinking is influenced more by their own knowledge associations than by conventional logic. Logic means less to children than to adults. It would appear that children are satisfied as long as the search is efficient and easy.

Table 3. Relations between keywords in the children's information architecture.

Relation between keywords	Children's IA (%)	Adults' IA (Yahoo directory) (%)
Related	18	0
Inclusive	50	89
Illogical jump	24	11
Inversely inclusive	4	0
No relationship	1	0
Identical	3	0
	100	100

With regard to the second characteristic of ambiguous categorisation, we found that the ambiguity is not simply a matter of human subjectivity but due to the ambiguity of the language itself. It is reasonable to hypothesise that children with underdeveloped language skills are more likely to have an ambiguous knowledge structure than adults, and this hypothesis may account for errors in the children's information architecture.

With regard to the third characteristic, namely that children produce a wider and shallower information architecture than adults, we found, as shown in Table 4, that the information architecture created by children is, on average, 33% shallower and 70% wider than the adult information architecture of Yahoo Kids. The reason for this outcome is that children lack the ability to organise their ideas systematically. In the experiment with Info Tree, children were given a topic and asked to use radiant thinking (Buzan and Buzan 1996) to generate related ideas. In the experiment with Info Block, the children were given a set of keywords and asked to categorise them until the words converged to one general idea. We observed, however, that the children often stopped structuring once they had put related words together. Hence, the lack of logical criteria results in horizontal growth rather than vertical growth in information architecture.

Table 4. Comparison between the information architecture of children and adults.

	IA created with Info Block	IA created with Info Tree	Yahoo Kids directory
No. of info units	55	68.1	45
Width
1st level hierarchy	7	9.4	6
2nd level hierarchy	25	28.3	24
3rd level hierarchy	38	28.8	7
4th level hierarchy	2	6.0	8
5th level hierarchy		2.0
Depth	3	3.6	4
Avg. breadth (no. of info units/depth)	20	19.6	11.3

5.4.2 Evaluation of participatory design process

(1) Tacit needs. In evaluating the way children produce information architecture, we can see a close relation between the tacit needs and cognitive processes of children. The use of verbal protocols to identify the usability problems may be less effective because of children's inadequate language skills. Observation of children's behaviour may disclose the problems, though researchers must be able to identify the causes and understand the cognitive characteristics of children. Our toolkits were designed to help children build their own information architecture. The information architecture created by children reflects a cognitive level that differs from the information architecture of adults.

(2) The role of the moderator. Each team's results varied in quality and quantity. During the experiment, we discovered that the results were often influenced by not only the type of toolkit but also the atmosphere of the team. The results varied significantly among the teams that used the same toolkits. We also observed that the success of the participatory design process depended on the role of the moderator. A moderator in a successful group showed the following characteristics.

	He or she actively participated in the process to prevent children from digressing; they also made sure that every child was involved and that the result was not overly influenced by the opinions of the group's leaders.
	He or she treated the children objectively. If a moderator treats children as inferior because they are young, the children seem to feel underestimated. On the other hand, if he treats children as peers, the children seem to feel awkward.
	He or she paid attention and helped the children to quickly identify and solve problems. For example, by ensuring that the group was not dominated by one or two outspoken members of the group, as such dominance tends to isolate or discourage others. The group had to be facilitated so that everyone participated equally and maintained an interest in the task.

(3) Pros and cons of video recording. The advantages and disadvantages of using a video camera to observe children are well known. Druin (1998) recommends note-taking rather than videotaping for a contextual inquiry because note-taking inhibits the Hawthorn effect: that is, the tendency of children to ‘perform’ in front of a video camera. Iverson (2002), on the other hand, claims that video cameras may stimulate children and help observers to uncover needs that are normally not observed, and suggests that a video camera can be used as an instrument of provocation. In our experiment, some of the girls from grades five and six refused to be filmed on camera, but the camera appeared to have no significant consequence on the rest of the participants. The children did perform when they first saw the camera but they soon became absorbed in the tasks and ignored its existence. However, we used note-taking for the girls from grades five and six who refused to be filmed. Aside from these cases, the video camera was an effective tool for collecting rich data about the children.

(4) Fun. ‘That looks like more fun!’ This comment was uttered by a child in an Info Block team while pointing at the Info Tree. Our observation of the children confirms that they expected each task to be fun. In other words, the children considered each task to be a form of play, not work. Fun is important because it keeps children focused. The recommended testing time for children is 45 min, but children can stay focused for longer if they find the tasks to be fun (Hanna et al. 1997). The element of fun therefore improves the quality and quantity of the test data. Although the children from grade three and four soon became absorbed in the tasks, at times the sixth-graders and some of the seventh-graders appeared bored; for example, they would play independently with the toolkits and digress during the test. This behaviour suggests that the toolkit may have been inappropriate for children on the cusp of the formal operational stage of cognitive development.

(5) Effectiveness of questionnaires for testing children. After the prototyping session, the children were asked to discuss with moderators any aspects or problems they faced during the experiment and to make a general evaluation of the toolkits. The moderators led the discussion by asking the children the following questions: (1) Which keywords were hard to group and why? (2) Is there any illogical decision in the grouping and why? (3) What criteria did you use in grouping the keywords? (4) How did you choose the main keyword? (5) Do the main keywords properly represent the group? If not, why not?

The children gave detailed answers to concrete questions on issues such as demographic data. However, they responded with short statements such as ‘I just did it’ or ‘yes’ or ‘no’ to questions on information architecture, probably because the topic was too abstract and difficult for discussion. The children were very honest and kindly told us what they knew about each topic but, in spite of their positive attitude, the discussion was hindered by their level of cognitive development.

(6) Cooperation. At six years of age, the concrete operational stage, children can play cooperatively and are beginning to learn social and reciprocal activity from one another (Yoo 2000). The children in our research showed a tendency to perform tasks individually, especially in the task of generating ideas. Because our goal was to measure the collective generativity of children, and not individual capabilities, teamwork was essential for the generation of ideas and keywords. Furthermore, the moderator was entrusted with the task of reconciling any disputes and finding the most representative opinion of the group. As with groups of adults, there were several outspoken leaders among the children. An impulsive child is a faster decision-maker but tends to make more errors than a thoughtful child. Some of the younger children may be intimidated by the older ones and become isolated from the decision-making process. In all cases, the moderator should facilitate the group and give everyone an equal chance to contribute ideas.

(7) Education. While manipulating the toolkits, children learn to organise information. A girl from grade five, for example, asked her friend if ‘sea’ belonged to ‘vacation’ and vice versa. After a short discussion, they concluded that ‘vacation’ could go with other places such as ‘mountain’ or ‘river’ and agreed to put ‘sea’ under ‘vacation’. This incident suggests that the task of building information architecture helps a child learn the relations between concepts or ideas. Moreover, while performing the task, a child acquires knowledge because knowledge pertains to the acquisition of information and the comprehension of relations. Working in teams stimulates competition and cooperation between children and helps them develop social skills.

6 Discussion

On the basis of the experimental results, we have made several suggestions for designing information architecture for children.

In designing information architecture for children, we should determine the degree of logicality for an age-appropriate level of cognitive development. Hence, the depth of information architecture should not be sacrificed for logicality. Information architecture created by children is wider and shallower than the information architecture of Yahoo Kids, indicating that the categorisation process of children tends to be less systematic than that of adults. As mentioned earlier, the thinking of children, particularly with respect to the creation of information architecture or the navigation of websites, is determined more by knowledge association than by conventional logic. This phenomenon implies that information architecture designed with an adult level of logic may lead to inefficient navigation by children. Thus, the participation of children is essential in the design process.

The children's ambiguous categorisation of the contents left some information belonging to more than one category. In several cases, children provided multiple routes to a single destination. For example, ‘Tchaikovsky’ was located under the ‘Encyclopedia’ category as well as under the ‘Classical music’ category. In other words, children's information architecture was more flexible than adults'. According to Rosenfeld and Morville (1999), navigation systems should have balance in flexibility and context. If a hierarchical navigation system lacks flexibility, users are forced to move up and down the tree structure of content hierarchy and navigation becomes inefficient. On the other hand, if too many hyperlinks exist in a hierarchical navigation system, the system becomes too complex and users are likely to get lost in the system. In the case of children's information architecture, the system tended to be more flexible than adults'. Because children are less influenced by conventional logic and tend to think in a more creative and flexible way than adults, we recommend that information architecture for children should be more flexible than one for adults. One way to improve flexibility is to provide cross-listing and associative links (Rosenfeld and Morville 1999). They suggest that navigational flexibility can be improved by the use of cross-listing, i.e. by increasing the number of routes to a destination, even if the clarity of content is diminished. At the same time, associative links can facilitate navigation by providing children with a set of keywords associated with a topic.

7 Limitations of the study

This study is based on the assumption that the children at the age from 8 to 12 understand the notion of hierarchical information structure and that they can make one for themselves. We verified this assumption by conducting a pilot test and in-depth interview prior to the experiment. Besides, the fact that navigating a web directory is a reverse process of building hierarchical information structure and that children use websites such as Yahoo Kids without difficulty also implies that our assumption is valid. However, most participants in the experiment were familiar with mind maps and had experience of visiting Yahoo Kids, which we believe helped children understand the experiment better. It is not known whether knowledge of mind maps or experience with web directories were important factors for completing these tasks. Neither do we know how to explain the concept of information hierarchy to children who lack such knowledge and experience.

Our future study will be on validating the suggestions we proposed in the discussion session. We plan to design a website based on children's information architecture and compare it with Yahoo Kids by conducting a usability test.

8 Conclusion

A participatory design approach to information architecture design for children is effective for determining the user needs because it enables designers to look at problems from a child's standpoint and, simultaneously, deals with childhood traits such as shyness and immature language skills. Rather than evaluating the usability of information architecture with children as testers, we let children design information architecture for themselves as designers and analysed the result. We developed two cognitive toolkits with which children created information architecture and the output was compared to the information architecture of Yahoo Kids directory. The comparison revealed that the information architecture of children reflects their cognitive processes; children's information architecture is generally shallower and wider, and its content is more ambiguous and contains more logical errors than that of Yahoo Kids directory. On the basis of our results, we recommend that the logicality of information architecture for children should be determined in accordance with the cognitive development of users.

Notes

1. Collective generativity is defined as stakeholders' collective ability to generate or create what they think or imagine.

2. Info Block and Info Tree are the trademarks protected by copyright.

3. Smilansky (1968, cited in Yoo 2000) divided play into ‘construction play’ and ‘socio dramatic play’. In construction play, children create things with various materials such as blocks.