|
Article Excerpt
In the Playground project, we applied a constructionist learning perspective in order to build a computational learning environment in which children could design and build their own video games. In this paper, we present results from a study where children were given semi-structured programming tasks in an adventure game designed to investigate their understanding of program mechanisms. We analyze two children's solutions and approaches to a task as a matter of adaptation of talk and actions to different perspectives involved in the ongoing discourse. The establishment of a common perspective between child and investigator throughout the work sessions proved to be central to how the children approached their work. The analysis showed that in order for children to learn to understand how mechanisms that control a game work, they must learn to adapt their perspective to the expectations of each subtask and to the task as a whole. We show how one child is able to see the expected perspective in each subtask, whereas the other child finds that this is much harder. The support given by the investigators was also of great importance in facilitating these processes.
**********
INTRODUCTION
Constructionist perspectives on learning and education argue that one way to achieve powerful learning is to have children collaboratively build and design their own artifacts, which they share with others, e.g., friends, parents, and teachers. One such construction activity is computer programming, which has been thoroughly researched within the Logo-tradition. Programming in Logo and other languages has proved to help children learn about complex phenomena, for instance, within physics (diSessa, 2000), and fractions in mathematics (Kafai, 1995). Smith and Cypher (1999) argue that the kind of problem solving in which children become engaged through programming tasks, supports them in forming hypotheses about why things happen, adjusting properties of the task, observing the results, and analyzing the outcomes. This is supposed to resemble "scientific" ways of thinking. Furthermore, Bruckman (2000) has argued that if such activities also involve aspects of children's culture, this will be a scaffold for their motivation to dig deeper into topics that are central in domains such as mathematics or mechanics.
In the Playground project, we applied a constructionist learning perspective in order to build a computational learning environment in an animated programming environment called ToonTalk (Kahn, 1999). The Playground environment is designed for children ages six to eight to design and build their own video games. The games work as entry points for children to explore and learn about mechanisms that control formally defined systems such as video games.
The general goal of our research is to develop an understanding of how children learn in technologically rich situations and to develop ways to support children in such situations. In order to achieve our goals, a deeper understanding of how children use complex technology and appropriate it into their learning is required. The particular learning goal of the study presented here was that children, through composition, decomposition, and reprogramming of games and game components, would develop an understanding of the relationship between the programs they build, and the way those programs are played out as behaviors in a game. The ability to understand such relationships and mechanisms is particularly central to thinking within physics and mathematics (Noss and Hoyles, 1996; diSessa 2000), but also in other areas such as history (Hallden, 1997). Our earlier studies show that young children (six to eight years) can build meaningful things in a complex programming environment (Tholander, Kahn, & Jansson, 2002). An important follow-up to those results is to increase our understanding of what children understand and learn from such programming activities. Here, we present results from a study where children were given semi-structured programming tasks designed to investigate their understanding of mechanisms in an adventure game.
Another way to describe what we are studying is through a distinction between surface and deep-level features of computer software. Surface-level features refer to the specific instantiations of programming objects representing characters in a game or in a simulation. Deep-level features refer to the program code that determines the behaviors of the objects. The relationship between these two levels determines the view of a running piece of software. The goal of the work presented in this paper is that the children should develop an understanding of this relationship. The task conducted by the children was based on the deep-level features of a familiar game but the game were instantiated in a novel way, i.e., given new surface level features.
Previous Work on Children's Understanding of Programming
Since the 1970s, attempts to build programming environments for children have been made; these started with Logo and Smalltalk. A substantial amount of research on children's and students' use and learning of programming languages has been conducted (Clements & Meredith, 1992). In these early days, programming was considered a paradigm case of problem solving and, therefore, the ideal arena for "improving the minds" of students (Eisenberg, 1997). Problem solving has traditionally been described as iteration over activities such as problem identification, problem decomposition, solving of sub-problems, and composition of sub-problems to a whole (Polya, 1957). One line of research has focused on studying the validity of those promises by investigating the occurrences of cognitive gains from learning computer programming, and in particular for problem solving and mathematical skills (Hoyles & Noss, 1992; Noss & Hoyles, 1996; Palumbi, 1990). Most of these studies focus on "if and when" programming skills transfer to other domains, such as mathematics or physics, or even problem solving in general. The results of those studies are inconclusive in many respects (see Palumbi, 1990) and do not establish any clear connections between programming and general problem solving skills. Pea & Kurland's (1984) often-cited studies showed no significant indications of cognitive gains from students' Logo programming. Hoyles and Noss (1996) extensively analyze the studies of Pea and Kurland and criticize their interpretation of the Logo-vision as suggesting that Logo programming per se would lead to cognitive gains. From Hoyles' and Noss' point of view, it is quite clear that productive use of Logo always has to be accompanied by a rich culture of learning. If one wants to study students' learning with Logo, one cannot do this without thoroughly considering the culture of learning that surrounds and is created by the Logo activities. Furthermore, Noss & Hoyles find it unfortunate that the major impact of Pea's & Kurland's studies has been their conclusion that programming does not lead to the expected cognitive gains, and should, therefore, not be used in schools for such purposes. What is often left out is that Pea and Kurland also concluded that usages of programming in school settings have beneficial effects given a broad focus not only on the programming tools but also on how to support teachers and students to create a productive culture of learning around the programming activities (Noss & Hoyles, 1996). A related line of research in students' programming concerns the analysis and systematization of students' difficulties and misconceptions in learning to program (see Confrey, 1990). Resnick (1990), as an example, studied 5th grade children's' understanding of concurrency in MultiLogo programs (a precursor to StarLogo). He identified three different "bugs" in the students' reasoning (problem decomposition, synchronization, and object-oriented bugs), and tried to establish how these deviate from an expert's model of these constructs. Similarly, Rader, Brand, & Lewis (1997) studied children's understanding of the programming elements in StageCast Creator. The results suggested that children seldom explored the underlying programming model and mostly performed routine actions. Similar results were found in a study by Gilmore, Pheasey, Underwood, & Underwood (1995) that indicated that the depth of children's understanding of the involved programming concepts was limited. This mirrors a theme of most Logo studies, namely, the goal of using programming as a means to get learning effects that can be transferred to, and used in other domains. Most studies of this kind have adopted a view on transfer that tries to find effects pre-determined by the researcher, such as problem solving or a deeper understanding of programming concepts that go beyond children's actual programming activities (Koschmann, 2002). Even though programming still is often seen as a paradigm case of problem solving in the traditional sense, this paper argues that such an account of children's activities is not sufficient to understand what children in lower age groups learn and understand from programming activities. Moreover, the problem solving and cognitive gains approaches have limited possibilities in providing constructive input to the development of the scaffolding that children might need when learning with programming tools. Clements and Meredith's (1992) review of a considerable amount of the Logo research points to varying results with respect to problem solving. However, a common theme of the successful projects, in terms of children's learning, was that they all had a high degree of teacher involvement; the teachers served as mediators helping children to make connections to other domains and general problem solving principles.
An important aspect with respect to our research is that most of the Logo research rather narrowly focused on children's ability to develop an understanding of programming language terminology such as looping, iteration, or recursion and how such understanding transfers to problem solving in general. Most studies failed to show that children developed an understanding in terms of programming language terminology or any transfer of such understanding to new domains.
Several of the Logo studies (Clements & Meredith, 1992, Clements, 1995) and the KidSim studies (Rader et al., 1997, Gilmore et al., 1995) noted great enthusiasm, enjoyment, and enriched social cultures among the children. However, such issues were seldom focused upon but rather noted as side effects of the phenomena in focus. The theoretical and methodological approaches in the studies discussed above did not provide for any understanding of children's meaning making and their actions in the situations as they unfolded.
Socio-cultural Views on Misconceptions and Transfer
Most of the research on student programming and construction as discussed above focused on identifying the bugs and problems they had in learning to program. Similar work has been conducted in science education where substantial attention has been paid to identifying children's misconceptions (or alternative frameworks or naive theories) in understanding scientific phenomena (Caravita & Hallden, 1994; Chi, Slotta, & de Leeuw, 1994; McCloskey, 1983; Vosniadou, 1994) and in finding ways to overcome these. A common theme of that research is the assumption that children's interpretations of the phenomena under investigation are relatively stable across contexts and situations, and thus only subject to influence by the circumstances of the particular experimental situation (Schoultz, 1998). Schoultz, however, challenges the assumption that it is possible to set up studies, which establish "a neutral ground" (Schoultz, 1998, p. 62) where children's understanding of phenomena can be studied and stipulated. Instead, by replicating well-accepted experiments, Schoultz showed that children's explanation of scientific phenomena was largely determined by discursive circumstances, and can therefore not be used to determine fixed (mis) conceptions (1).
Furthermore, the research on the relationship between cognitive gains, problem solving, programming, and the understanding of science strongly connects to the debate regarding how to understand the notion of transfer (Anderson, Reder, & Simon, 1996; Andersson, Reder, & Simon, 1997; Carraher & Schliemann, 2002; Greeno, 1997; Lave, 1988). The view on transfer taken in this analysis is in line with the socio-cultural viewpoint that although research rarely succeeds in demonstrating the existence of transfer, it is clear that learners draw heavily on prior experience and knowledge when acting in new situations. However, learners do not always utilize the knowledge or the experiences that the researchers expect or wish. Thus, the consequences of the failure of transfer studies is not that transfer does not exist, but rather that we need to conceptualize and study transfer in new ways (Lave, 1988). Carraher & Schliemann (2002) even claim that the study of how people bring knowledge and...
|
