1 Young Children’s Learning while Building Water-Flow Systems Sharona-Tal Levy Tel-Aviv University, Israel 972-3-6407799 (work), 972-8-9454220 (home) Email: [email protected] ABSTRACT One of the fundamental ways of learning involves the cycling between constructive action upon reality and reflection on its results. This study examines such learning processes among kindergarten children, who are building dynamic water systems. Questions regarding the knowledge obtained and its form of acquisition are addressed. The development of the children’s understanding of content principles and different temporal patterns are investigated over a time period. A water-flow system was developed and used as a half-open learning environment, where the children (8 girls and 7 boys, 5-6 years old) built individually during four sessions. Interviews were conducted before and after the building period, and at the end of each session. The pretest and posttest were also administered to a control group, who did not build any systems, but underwent alternative treatment concerning mythology and astronomy. Results of the investigation are discussed along the following lines: (a) The overall learning through building. (b) Processes and indicators of change: Obtaining and losing consistency in the processes of understanding complex systems. INTRODUCTION Technology education offers a powerful way of learning for young children. Through building artifacts and discussing how they work, a great potential for learning can be seen (Piaget, 1956). Making technological products has been viewed as an important channel to understanding and perhaps appreciating the knowledge upon which they are based. The making of objects that work involves one of the fundamental ways of learning, the cycling between constructive action upon reality and reflection on its results. This study attempts to uncover the processes taking place in the children’s changing understanding of technology systems, while building them and increasing their own proficiency. In an attempt to disentangle the different mental processes participating in “learning while building”, one can separate the following strands: (a) Motor action – the physical motions of construction, which change and shape the structure. (b) Perceptual information extracted from reality, channeling information from the structure into the mental system. (c) Rule-models or concepts - the causal relations between particular configurations and ways of operation on system behavior. (d) Problem-solving strategies.

2 The following conceptual map names and relates the mental processes that participate in building artifacts.

The current research performed investigates three parts and their relationships: concepts, actions and perception. This paper describes one of these - the changes in rule-models while building water-flow structures. This study concerns young children and their learning through building unfamiliar water-pipe systems. Younger children’s causal models of physical phenomena are usually focused on a single causal dimension (Siegler, 1978; Case, 1989). On the other hand, experts build more complete representations of the problem than novices because of the extra knowledge they have available. This can be seen among young children in a comparison between expert and novice 5-7-year olds’ knowledge of dinosaurs (e.g. Chi, Hutchinson & Robin, 1989). In this investigation, the children start out as novices in the content of hydrodynamics, but are expected to develop some level of expertise beyond the initial level. The research questions we pose are: What characterizes children’s rulemodels of water-flow in pipe systems? What changes are seen in these models over a time period when they are building such systems? METHOD The sample included 29 children, 15 girls and 14 boys, selected out of 80 children in an Israeli middle class public school, and randomly assigned to experimental and control groups. The children’s ages spanned 5’2”-6’3” with a mean age of 5’8”, SD=3”. Two sets of instruments have been developed. One is a construction kit for building large water-flow systems. It is modular and transparent, and its components enable the creation of various systems. One can control the water flow using diverse components (pipes, faucets, vessels, connectors and qualitative speed measuring devices) to determine the relationship between the streams’ features and the following variables: height, exit-hole cross-section, resistance, hierarchical structure of the system and the system’s water flow control. The second is a series of tasks using this construction kit as a half-open learning environment, with the children building different systems. The tasks were designed as a progression of increasing complexity. E.g., the second task was to create a plumbing system for a model house with two stories, so that neither neighbor would complain the other was getting more water.

3

PROCEDURE The experiment took place during six meetings spaced slightly over one week apart. The first and last sessions were a pretest and a posttest. The children built individually during four sessions. Interviews were conducted at the end of each session. The pretest and posttest were also administered to a control group, who did not build any systems, but underwent alternative treatment involving mythology and astronomy. The building sessions were videotaped. The interviews included questions, aimed at eliciting the children’s rulemodels regarding water behavior in pipe-systems. The questioning ran along the following lines: a real system is presented, described, its parts and routes are emphasized. The child is asked whether a particular variation/s which is pointed at makes a difference for the water streams, and if so - what kind of a difference. S/he is asked to describe the water streams for the two states of the varied feature, and explain this prediction. The same question is repeated later in the interview using a picture of the system. Here, the child is asked to complete the streams in the picture, describe them and explain why they are so. The following variations were employed: pipe-exit height, hole-width and resistance. Single-variation tasks were used as well as dual-variations, usually in a compensating relationship. 33 tasks were administered during 6 interviews; each performed 3 times in a session (real system, drawing, explanation of picture).

Factors determining water-flow from pipe

Double-variation task: pipe-end height (higher pipe-end – less flow) and resistance to end (longer pipe - more resistance – less flow).

4 RESULTS The children’s predictions and explanations in the different tasks were recorded, coded and analyzed as rules: If (variation, direction) then (water streams description and comparison). What is learnt? The children’s rule-models were coded using the following scale: 1.

No rule.

2.

Single rule, incorrect

3.

Single rule, correct.

4.

Two different single rules provided in different contexts, at least one incorrect.

5.

Two different single rules provided in different contexts, both correct.

6.

Two rules, at least one incorrect.

7.

Two rules, both correct.

The scale is based on Siegler’s (1978) description for rule-models, with one addition: levels 4-5 reflect a transition stage, showing instability, while vacillating between two different and relevant single rules in different contexts of the same tasks. The placement of this response type on the scale is determined from the data, as will be shown. In the pretest, no group differences are found. For height-variation and holevariation, the median is the incorrect rule. For resistance-variation, the median is the correct rule. The following box-plot graphs1 describe the posttest rule-models for the two groups.

Posttest rule models group comparison:

Posttest rule models group comparison: Double variation tasks

Single variation tasks 8

6

Rule model

6

4

5

Variation

3 2

Height

1

Hole-width

0 N=

Resistance

15

15

15

14

Experimental

14 Control

Group

1

Rule model

7

5

14

Variation

4 3

Height & resistance

2 1 0 N = 15

Hole & resistance Height & hole-width

15

15

14

Experimental

14

14

Control

Group

The thick line represents the group’s median. The top and bottom lines in the rectangle show the 75% and 25% quartile.

5

Tables 1, 2: Single- and Double-variation tasks rule-models and statistics.

The experimental group’s rule-models improve, but not the control group’s. The final rule-models are correct for the single-variation tasks, but they are not integrated into a multiple-variation consistent model. Although the children show an ability to integrate two variations into a single explanation, this is not a general change.

6 What makes a difference for water flow? The conditional “if” parts of the children’s rule-models are examined for the causal status of the different system features in determining flow characteristics. The proportions of these features in the single- and the double-variation tasks’ responses are compared with the actual task variations. The difference between the two is termed the “deviation from task-variations”. The chi-square test statistic was used to calculate this deviation, and to examine the deviations for significance. These statistics are viewed over time: Experimental group deviation from task variations as a function of time - single and double variation tasks 60

Chi-square test statistic

50

Single variation Double variation

40

30

20

10

0

Pretest

1

2

4

Posttest

Session

One can see the children’s causal models of water-flow in pipes going through two interesting processes: (a) Increasing correspondence between task complexity and the children’s model complexity. (b) Intermediate strong deviation from task variation arises. It is due to a bias towards pipe-end width, as a single source of variation in water-flow. (c) The latter is not dependent on building, as it was observed for the control group too. How many rules? A rule structure denotes the number of rules and their stability, disregarding their correctness. Both an incorrect height rule and a correct height rule are “single rules”. The other structures are “fluctuating single rules” and “double rules”. Their timelines are presented:

7

Rule structures in double-variation tasks

Proportion of rules (%)

100 90

Single rules

80

Fluctuating single rules

70

Double rules

60 50 40 30 20 10 0

Pretest

1

2

4

Posttest

Session

In the temporal progression, single rules are dominant throughout, but are gradually displaced, first by 2 fluctuating single rules, and then by double rules. Thus the shift from simpler to more complex rule-models supports the proposed progression: single rule, fluctuation between two single relevant rules and double rules. The biased response and the fluctuating rule pattern were examined within individual progressions. Most of the children (60%) who exhibited a bias towards hole-width did so before vacillating between single rules. The fluctuating response pattern is an intermediate unstable phase, observed usually before attaining more complex rule-models, but also before more correct, but not more complex rule-models. In addition, a ‘balancing’ response pattern was found for 60% of the children, where irrelevant hole-width rules are provided together with the relevant rule. Comparison of this response type with the biased response shows them to be distinct, as the first peaks much later than the latter. Children are reasoning mainly with single rules, even in double-variation tasks. During the building period, a biased response is first to appear - when the children focus on a single feature as explaining the water-flow. Following that, a fluctuating response pattern is seen when different relevant rules are provided in different contexts. At the end, the dominant rule during the bias is combined with the relevant rule when reasoning about water behavior, and all the children can provide at least one double-rule. Simple and compound tasks It has been found that most of the children held correct models for the resistance variable. Therefore the separate height rules and the hole-width rules are examined for the time of conceptual change in the various tasks and then compared.

8 Both group and individual results show a surprising pattern:

Hole-width rule model

Hole-width rule model

in Hole-width variation tasks

in Hole-width & resistance variation tasks 6.0

Hole-width rule model

Hole-width rule model

6.0 5.0

5.0

4.0

4.0

3.0

3.0

2.0

2.0

1.0

1.0

0.0

0.0

N=

15

15

15

15

15

Pretest

1.00

2.00

4.00

Posttest

N=

14

1

15

15

Pretest

1.00

4.00

Posttest

Session

Session

Hole-width rule in single and double variation tasks

Height rule in single and double variation tasks

16

16

14

Single > Double

12

Double > Single

10 8

Consistent

6 4

Number of children

Number of children

14

Single > Double

12

Double > Single

10 8

Consistent

6 4 2

2

0

0

Pretest

1

2

4

Posttest

Session

Pretest

1

2

4

Posttest

Session

In many cases, one can see a shift to the correct rule first in a compound task, which includes a variation for which the children hold a correct rule - and only later in the simple task. The pattern is less clear for the hole-width rule. The same does not occur for the height and hole-width tasks, when both rules are incorrect at the start. DISCUSSION Complexity: technology versus cognition Technological systems which surround us everyday are complex objects. They contain many parts, are organized in different structures, and utilize a manifold of causal rules - being the fruit of many years of human-effort in accumulating and generating knowledge. The question that we pose here is: how does a single human, a young child in this case, approach this complexity, disentangling or constructing it in reality and in understanding. When novices, the children use a single rule to explain system behavior - be it simple or complex. Similar response types are seen in different physical (Siegler, 1978) or technological (Levy et al, 2001) tasks.

9 Following a period when the children are building such systems, complexity in reasoning tends to follow complexity in the systems, though not completely or consistently. At the end of the building period, the children are all explaining water-flow in the pipe-systems with complex rules. Nevertheless, this change is not general. The activity of building complex technological systems is seen to encourage children to go beyond their everyday abilities, as new rules are discovered, tuned to reality and integrated, finally replicating the causal structure of the artifacts they have been building. When comparing responses to simple and complex tasks, a surprise was in waiting. We usually expect learning to occur first in simpler settings, and only later applied in complex settings. What has been found, is that the contrary may occur. Not simpler systems, but more complex systems, where one rule was previously known, are those where learning of new rules occurs first. A possible explanation for this lies in the reinforcing character of correct rules. Model revision doesn’t occur when a cognitive system is confronted with a totally confusing situation, even if simplified. In order to facilitate learning, the unfamiliar needs to be paired up with familiar phenomena, even if this means ‘complicating’ the situation. Anchoring and then losing consistency on the way to achieving higher complexity in reasoning As building progressed, the general trend was of increased coherence with task variations, associated with a partial shift from single rules to the more complex interacting double-rule explanations. The builders’ ability to encode and in some cases coordinate two dimensions increased throughout the building period. A learning progression is proposed where a more robust rule model is achieved through the sequential enhancement of three abilities: 1.

Extending a rule model’s applicability or increasing the consistency of its use.

2.

Encoding multiple features.

3.

Rule integration.

It is suggested that there are two factors encouraging these changes. One is the effortful building activity that encourages a decrease in the number of operations to solution by increasing the power of prediction. The other is the interview situation, which motivates the children to reflect on their responses. Expenditure of effort in encoding more system properties and in the coordination of a larger number of rules is compensated for by economy in action. Preliminary models: isolated single rules Prior to building, any variation in system features was seen as changing water-flow. This is seen in the high degree of fit in the single-variation tasks. The differential causal status is seen only in the double-variation tasks, since usually only one rule was provided. Therefore, although the children provided rules regarding all system feature variations, these were isolated rules, in the sense that they cannot be mutually coordinated. Increasing consistency: anchoring reasoning processes

10 The picture changed in the next interview. After the first building session and peaking in the second session, a rise in irrelevant hole-width rules was seen, with the children focusing on a single system feature, ignoring the others. This bias was transitory and later diminished. It is claimed that before tuning into the complexity of the multiple variations, this temporary biased phase is necessary. Creation of consistency in reasoning is the first stage in rule model progression - anchoring the fleeting variations and their outcomes, by strengthening the strongest rule-model. Only then, can the individual detect conflicts between predictions based on this model and reality. Thus their accomplishment at this stage is consistency, an anchor in a sea of variation. Encoding multiple dimensions - and losing consistency The consistency achieved did not satisfy the builders. Through building, the children are initiated into more complex situations, where a single rule cannot successfully predict system behavior. The next step is an increasing awareness that the explanations for water flow in their structures must involve more information. In order to increase the power of their models in guiding the building process, additional system properties need to be attended to. At this phase, additional system properties are viewed as causal, but not at the same time nor integrated into a single model. The children fluctuate between the previously preferred rule and another task-relevant one, inconsistently providing them in different contexts. During this phase, encoding capabilities are advanced. Encoding is described as noticing potential explanatory variables. In learning, formulating new rules is preceded by the recognition that previously unattended dimensions may be relevant to the task or that their variation accounts for observed outcomes (Siegler & Chen, 1998). At this time, the formerly un-encoded features are now encoded and used in reasoning, but in an unstable manner. The price paid for increased encoding is decreased consistency. Coordinating multiple dimensions in double-rules After the children had enhanced their reasoning uniformity, they decreased it in favor of fine-tuning to more system properties. Still, they were not satisfied. Although they fluctuated between an increased number of dimensions, they had not yet integrated them in a single stable model. The need for consistency pushes the cognitive system beyond this transitory stage to more complex reasoning, which both encodes the causes and integrates their outcomes. This can be seen in the double rules, which were provided at higher frequencies in the later building sessions. After both consistency and encoding have improved, a robust model where integration of the two encoded dimensions in a double-rule can be achieved. CONCLUSIONS In summary, building complex technological systems encourages learning of new rules and their integration into consistent models. This learning is facilitated when an unfamiliar variation is coupled with a familiar reinforced

11 variation, rather than in a simplified situation where only the unfamiliar is varied. The complexity of reasoning is reached through a process of anchoring reasoning and increasing its consistency, and then by losing this consistency in favor of noticing more features. REFERENCES Case, R. (1989) ‘Science teaching from a developmental perspective: The importance of central conceptual skills’ Chapter 8 in Adey, P. et al (Eds.), Adolescent development and school science, The Falmer Press. Chi, M., Hutchinson, J.E., & Robin, A.F. (1989) ‘How inferences about novel domain-related concepts can be constrained by structured knowledge’ Merrill-Palmer Quarterly, 35(1), 27-62. Levy, S.T., Mioduser, D., Talis, V. (2001). ‘Concrete-abstractions stage in kindergarten children’s perception and construction of robotic control rules‘ PATT-2001 conference: New Media in Technology Education, Haarlem Holland. Piaget, J. (1956). The Child’s Conception of Physical Causality. Pp. 233-234. Littlefield, Adams and Co. Siegler, R.S. & Chen, Z. (1998) ‘Developmental differences in rule learning: A microgenetic analysis’ Cognitive Psychology, 36, 273-310. Siegler, R.S. (1978) (Ed.) Children’s thinking: What develops? Hillsdale, NJ: Erlbaum.