Proc. of ICCE2006, pp.535-542(2006)

Property Exchange Method for Automatic Generation of Computer-Based Learning Games Takanobu UMETSUab, Tsukasa HIRASHIMAb, Akira TAKEUCHIa a Computer Sicence and Systems Engineering, Kyushu Institute of Technology, Japan b Graduate School of Engineering, Hiroshima University, Japan [email protected] Abstract: Learning games are very useful to realize highly motivated learning. Developing effective and high-motivating learning games, embedding learning materials into a game is a promising approach. As a concrete method to embed learning materials into a game, we propose Property Exchange Method in this paper. The method can deal with card games where players manipulate cards following their rules. The rules usually only deal with limited attributes of the cards. Therefore, if the learning materials possess the same attributes, players can manipulate the materials in the same way following the same rules. Since players can play the game with the same rule dealing with the same attributes, it is expected that the game keep the characteristics of the original game. In this paper, we also introduce a system that automatically generates computer-based learning games based on the method. An experiment use of the generator is also reported. Keywords: Design Methodology, Intelligent Learning Environment, Game-Based Learning, Edutainment

Introduction Motivation is one of the most important factors in learning. Many researchers of learning environments, therefore, pay special attention to learning games as a remarkable approach to realize highly motivated learning. However, making a learning game is not easy task because the developers satisfy both motivation and learning effect. Although there are several investigations for design methods of learning games, most of them are only a kind of guidelines that the developers have to keep [1-6]. The developers, therefore, are required to have enough knowledge and experiences regarding learning games. Based on these considerations, we are investigating a concrete and simple method to design a learning game systematically. In our research, we don't try to design a learning game from scratch. Instead, we have adopted an approach to transform an existent game into a learning game while keeping the characteristics of the original game. Developing effective and high-motivating learning games, embedding learning materials into a game is a promising approach. As a concrete method to embed learning materials into a game, we propose Property Exchange Method in this paper. The method can deal with card games where players manipulate cards following their rules. The rules usually only deal with limited attributes of the cards. Therefore, if the learning materials possess the same attributes, players can manipulate the materials in the same way following the same rules. Since players can play the game with the same rule dealing with the same attributes, it is expected that the game keeps the characteristics of the original game. In following chapters, we explain Property Exchange Method in more details. We also introduce a learning game generator developed based on the method. The generator generates a computer-based learning game from a computer-based game and a set of

learning materials. We have already generated several computer-based learning games for Trick-Taking and confirmed that they can be played as games, are useful for learning the mateirlas and keep the characteristics of Trick-Taking. 1. Design Method for Learning Games Many arguments about the process of developing a learning game have been discussed in order to alleviate the difficulty in the process. For examples, Malone suggested that a learning game should be designed that to be skillful in the game is valuable for the target learning [1]. Prensky also proposed that it is important to carefully decide the combination of a game style (which is adventure, action, sports, role-play, strategy or simulation) and a learning style (which is drill, questioning or observing) [2]. Klawe analyzed the role of navigation and interaction in a learning game [3]. Maragos insisted the role of multiple stages in a learning game [4]. Halff dealt with adventure games, in which the player assumes the role of a character in a fantasy world. He proposed that a learning game should restrict behaviors of a player in order to make the player learn in suitable order [5]. In ICCE2004, we proposed “Partial Exchange Method” to transform an existent game into a learning game by exchanging a part of the existent game for learning materials when the part is similar to the learning materials [6]. Although these design methods help developers design learning games as guidelines, most of the process of developing learning games left to a developer’s discretion. For example, in Partial Exchange Method, it is difficult to find an adequate set of learning materials and a part of existent game because there is no clear definition about the part or the similarity. Therefore, the developer is required to have enough knowledge and experiences regarding learning games. Based on these considerations, we propose “Property Exchange Method” as a concrete and systematic method. The method is a specialized method of Partial Exchange Method. Partial Exchange Method deals with all kinds of games, but Property Exchange Method deals with only card games. However, the limitation makes it possible to find an adequate set of learning materials and a part of existent card game systematically. Based on the method, we have developed a system that automatically generates computer-based learning games. We also report the results of experimental use of it. 2. Property Exchange Method 2.1 Concept Although a lot card games uses the same set of cards usually called “playing cards”, the plays of the games are different. Therefore, the play of a game is mainly decided by rules of the game. Then, a game is sometimes played in the same way with different kinds of sets of cards. For example, UNO can be played with either the playing cards or special set of cards for UNO. In other words, it is necessary only a part of properties of the card to play a card game. Based on these considerations, if we can prepare new set of cards that have enough properties that are required in the original game, we can generate a new game that is composed of the rules of the original game and the new set of cards. Then, it is expected that playing the new game is similar with the original one. This is the basic concept of the Property Exchange Method. Figure 1 shows the outline of the method. In this section, the method is explained in more details.

rule manipulate data type

data type

Cards that have the same or substitute data type card property property data type data type value value

learning target concept property property data type data type value value

exchange set of cards

set of learning target data

Figure 1. Model of Property Exchange Method First, we discuss what a card is. A card has a several properties. For example, a card called “seven of hearts” in the playing cards has “♥” as the mark (it is also known as suit) and “seven” as the number. “Seven of hearts” is a card name. “Mark” and “number” are property names. “♥” and “seven” are values of the properties. In the card games, the properties of a card are usually used with the properties of the other cards. Then, the way to use the properties are categorized into three, (1) distinguishing (the same or not), (2) ordering (less or grater, before or after), and (3) calculation. For example of Fan-tan, which is a major card game, the rules make players “put playing cards sequentially by number in each mark”. The players order the values of numbers and distinguish the values of marks. Therefore, the rules require cards to have characteristics that is “the values of the properties of the cards can be distinguished”, “the values of the properties of the cards can be ordered” or “the values of properties of the cards can be calculated”. Whether the manipulations of values can be carried out or not depends on the relations between values of the properties that have the same name, so that the characteristics are the relations. We call the relation “data type”. In other words, the rules require specific data type to the properties. Therefore, if a property has the same data type with another property, it is possible to exchange them without changing related rules. So, if we can prepare new set of cards that have the same data types with them of the original cards, we can make a new game by exchanging them without changing the rules. It is expected that the play of the new game is similar with the original one because the new set of cards that can be manipulate by the rules in the same way as the original cards. When the new card is composed of learning materials, the new game is a learning game of the learning. In the following subsections, the data type and the learning materials are explained in more details. 2.2 Data Type The manipulations of values can be carried out by using relations between values of the same property. We call the characteristics of the relations “data type” and categorize it into three. We call the attributes that can be distinguished “nominal scale”, the attributes that can be ordered “ordinal scale” and the attributes that can be calculated “interval scale”.

In nominal scale, values are only labels or names. If two values have the same name associated with them, they belong to the same category, and that is the only significance that they have. The only comparisons that can be made between variable values are equality and inequality. There are no "less than" or "greater than" relations among them, nor operations such as addition or subtraction. For examples, a mark of playing cards, color of card, name of elements in chemistry and name of countries are properties of nominal scale. In ordinal scale, values have order relations. Values represent the rank order (1st, 2nd, 3rd etc) of the properties measured. The natural numbers are ordinal. Comparisons of greater and less can be made, in addition to equality and inequality. However, operations such as conventional addition and subtraction cannot perform on this type. For examples, alphabet, ranking of production and experimental procedure are properties of ordinal scale. Interval scale has all the features of ordinal scale. In addition, equal differences between values represent equivalent intervals, so that differences between arbitrary pairs of values can be meaningfully compared. Operations such as four arithmetic operations can perform on values of this type. For examples, number of playing cards, electron numbers of elements in chemistry and years of historical events are properties of interval scale. The rules make players distinguish, order or calculate a value of a property. In other words, the rules require the property that has particular data type. If the rules distinguish values, the rules require the property has nominal scale, ordinal scale or interval scale. If the rules order values, the rules require the property has ordinal scale or interval scale. If the rules calculate values, the rules require the property has interval scale. For examples, the rules of Fan-tan, which is a playing cards game, is “put a card sequentially by number in each mark”. In this case a set of cards that has a property of nominal scale and a property of ordinal scale is minimum required. 2.3 Learning Style In learning, a learning material often has a name and several properties as well as a card of a card game. A property also has a name. A property has different value for each learning material. We usually have to memorize the values of properties of learning materials and their relations. For example, learners have to memorize values of properties of elements in order to check characteristics of the elements in chemistry. The element has properties, which are “period”, “group”, “electron number”, “atomic symbol” and so on. The characteristics of element depend on values of the properties, so that we memorize the values. Furthermore, the relations of elements should be memorized. For example, both lithium and sodium belong to the same group, so that they have similar characteristics. Memorizing the relation is also useful for checking characteristics of elements. We call such learning material “learning target concept”. One of activities for memorizing values of properties of learning target concept is checking the set of values, properties and a name of learning target concept many times. The activity makes learners associate values with the name of the learning target concept. For examples, we usually speak up the set many times, remind the set a number of times, and so on, for memorizing. As already mentioned, developers can generate a learning game that is similar with the original game by exchanging the set of cards of the game for the set of cards that are composed of learning target concepts when the learning target concepts have the data types that are required by the rules. In the learning game made by the exchange, players manipulate the values of the learning target concepts many times. The manipulating accompanies the process of checking the set of values, properties and a name of learning target concept, so that playing the game is useful for memorizing the learning target concept. In addition, the players distinguish, order and calculate the values of the learning

target concept, so that the players check relations of the values of the learning target concept. It contributes to memorize the relations. For example, the rules require players to “put an element sequentially by group in each period”. The players must check a set of values of groups and periods of element in order to manipulate them. The players also check order relations of the groups of the elements and equivalence relations of the periods of the elements in order to put the elements sequentially by group in each period. 3. Example of Property Exchange Method In this chapter, we explain an example of developing a learning game based on Property Exchange Method. First, a developer prepares information of a set of cards and learning target concepts. Information of a set of cards consists of “game name”, “card set name”, “card name”, “property name”, “minimum required data type for the property" and “value”. If the rules only distinguish the value, minimum required data type is nominal scale. If the rules order but don’t calculate, minimum required data type is ordinal scale. If the rules calculate, minimum required data type is interval scale. Information of a set of learning target concepts consists of “concepts set name”, “concept name”, “property name”, “data type of the property” and “value”. Table 1 and 2 shows examples of them. The rules of Casino calculate number of playing cards, so that require only a number property that have interval scale. Mark of playing cards is not described because the rules don’t manipulate it. Game Card set name name Casino Playing cards

Table 1. Information of a set of cards Card name Property Minimum name required data type Six of hearts Number Interval scale Seven of heats Number Interval scale

Value 6 7

Table 2. Information of a set of concepts Concept Concept Property Data type Value set name name name Historical Egyptian Year Interval scale 1952 events Revolution Place Nominal scale Arab Republic of Egypt Suez Year Interval scale 1956 Crisis Place Nominal scale Arab Republic of Egypt Second, the developer is required to find exchangeable pairs of card set and concepts set. When the data types of the concept set are as same as or substitute data type for required data types for the property of the card set, they are exchangeable. Interval scale is substitute data type for ordinal scale or nominal scale. Ordinal scale is substitute data type for nominal scale. For example, the rules of Casino require interval scale in Table 1, so that playing cards of Casino and historical events that have interval scale are exchangeable pair. Third, the developer exchanges the pair by exchanging the card set name for concept set name, the card name for concept name, property name of the card for property name of the concept that has required data type, value of the card for value of the concept. The name of the concept, property name of the concept and the value of the concept are written on each card. For example of the card is “Egyptian Revolution. Year=1952” in the case of Table 1 and 2. Forth, the developer generates a learning game by exchanging the new card set for the original card set. In the case of Table 1 and 2, players manipulate year of historical

events by calculating in the new game many times, so that the players can memorize the year of historical events and their interval. 4. Learning Game Generator In this chapter, we introduce a learning game generator based on the method. The experimental use of it is also reported. 4.1 System Specification

Learning game designer

Card game designer

Computer-based game (1) card (1) property

Set of concepts (1) concept (1) property (1) data type

(1) data type

(1) value

(1) value

(3) rule

(2) rule explanation for game

system Pair-finder

Card generator

(a) Set of cards for learning game

(c) Computer-based learning game

Explanation generator

(b) Rule explanation for learning game

Figure 2. System chart Figure 2 shows system chart. There are 3 kinds of outputs. When developer inputs (1) information of a set of cards and learning target concepts, the system outputs (a) information of new set of cards for the learning game. An example of (1) information is Table 1 and 2. An example of (a) information is “Egyptian Revolution. Year=1952”. When the developer inputs (1) information and (2) rule explanation of the card game, the system outputs (b) rule explanation of the learning game. The (2) rule explanation has to be text independent particular card by using the words “card set”, ”property”, “value” and so on. For example, when (2) rule explanation is “put seven of hearts, eight of hearts and nine of hearts”, the system cannot output. When (2) rule explanation is “put cards sequentially by value of property_1 in each value of property_2”, the system can output “put cards sequentially by value of year of historical event in each value of place of historical event”. If the developer gets the (a) information and (b) onto paper, players can play the learning game with printed cards following the rule explanation of the learning game. When developer inputs (1) information and (3) computer-based card game, the system outputs (c) computer-based learning game. The system makes (a) information of new set of cards by (1) information of a set of cards and exchange (a) the information of new set of cards for the data of cards of (3) computer-based learning game in order to generate (c) computer-based learning game.

4.1 Evaluations We conducted two experimental evaluations in order to access the system. One is an experiment for confirming the system can output various learning games. The other is for conforming the system can output learning games that have the same characteristics with the original games. In the first experiment, we tried to input (1) information of 5 games (Memory, Speed, Fan-tan, Casino and Trick-Taking) and 4 learning target concepts (elements, historical events, procedure of using microscope, English words) and (2) rule explanations of 5 games. Each set of cards of the 5 games was different data types, and the 4 learning target concepts were also different data types. When we input them into the generator, the generator output (a) sets of card and (b) rule explanations for 50 learning games. We confirmed that we could play the 50 learning games by getting (a) sets of card and (b) rule explanations onto papers. The results suggest that the system can output various and many learning games. In the second experiment, we have developed a computer-based Trick-Taking game, which is playing cards game on Windows. When we input the computer-based game and 3 learning target materials (elements, historical events, procedure of using microscope), the system output 13 computer-based learning games. We asked six subjects to play the 13 games and answer questionnaires. The six subjects often play Trick-Taking and two of six subjects are ranked player on Internet-Trick-Taking. The subjects suggested that they played each 13 games twice in order to understand the derivation of Trick-Taking. One play took about 5 minutes so that they play the 13 games for 130 minutes. The questionnaires are as follows. z Q1 was “How much do you know the learning target concepts?” 5. completely-known 4. much 3. so-so 2. a little 1. completely-unknown z Q2 was “Do you accept the activity of playing as a game?” 5. completely 4. almost 3. so-so 2. incompletely 1. different z Q3 was “Do you think that the learning game is useful for memorizing?” 5. very good 4. good 3. so-so 2. bad 1. awful z Q4 was “Do you think the learning game keep the characteristics of the original game?” 5. completely 4. almost 3. so-so 2. incompletely 1. different z Q5 was “Do you line the learning game?” 5. very much 4. much 3. so-so 2 a little 1. dislike The results are shown Table 3. Thirteen alphabets from A to M mean the 13 learning games. Numbers in the Table mean averages of the answers. The distributions of the answers are normal distributions. The results of Q2 and Q3 suggested that the generated games were accepted as useful learning games. The results of Q4 and Q5 suggested that most learning games kept the characteristics of the original games, but Game-E, -F -G, -H and -I didn’t keep the characteristics. We asked the subjects Q1 in order to investigate repercussions of knowledge about the learning target concepts on the results of Q2-5. The results of Q1 had no influence on other questionnaires, so that difficulty of memorizing has little concern with the learning target concepts.

Q1 Q2 Q3 Q4 Q5

A 4.7 5 4.5 4.2 4.7

B 4.7 5 4.5 4.2 4.7

C 4.5 5 4.5 4.2 4.7

Table 3. Results of Q1-Q5 D E F G H I 4.5 4.5 4.5 4.7 4.7 4.5 5 5 5 5 5 5 4.3 4.3 4.3 4.5 4.5 4.3 4.3 2 2 3.5 3.5 1.8 4.3 2.8 2.8 4.2 4.2 2

J 2.5 5 4.8 4.8 4.7

K L M 2.5 2.5 4.8 5 5 5 4.8 4.8 4.5 4.8 4.8 4.3 4.7 4.7 4.3

Table 4. Results of Q4, 5 after the adjustment E F G H I Q4 4.7 4.7 4.7 4.7 4.7 Q5 4.7 4.7 4.7 4.7 4.7 Closely examining the Q4 and Q5 results, we think that the rules and data types are important factors to provide characteristics of the game, but there are other important factors. We think the number of values and the distribution of values are changed in E, F, G, H and I, so that they didn’t keep the characteristics. Therefore, we adjusted the number and distribution of values in E, F, G, H and I as the original game in order to confirm the thinking. After the adjustment, we asked six subjects to play them and answer questionnaires again. The results are shown Table 4. The results suggested that if the developers can allow for the conditions of the number and distribution of values, it is expected to remove the problem. Although the subjects answered these learning games were useful for learning, we didn’t investigate the learning effectiveness of these learning games. However, the system could output the learning game that showed an effect on learning [6]. Under the circumstances, we think that Property Exchange Method can support designing learning games and the system on the method can generate computer-based learning game automatically. Although the system occasionally generates a learning game that doesn’t keep the characteristics of the original game, we can design another learning game that keep the characteristics from the learning game by adjusting the values of properties. 5. Conclusions In this paper, we have proposed Property Exchange method, which is a method to transform an existent card game into a learning game by exchanging learning target concepts for cards of the game while keeping characteristics of the game. The method only deals with card games. Instead the method is so concrete method that the application based on the method can generate computer-based learning games automatically from computer-based game, learning materials and their data types. We confirmed that the application could output computer-based learning games that keep characteristics of the original game. In future work, we should investigate factors that produce characteristics of a game in detail, because the application rarely output computer-based learning game that didn’t keep characteristics of the original game in the experimental use of the application. References [1] T.W. Malone (1981) Toward a Theory of Intrinsically motivating instruction. Cognitive Science, Vol.5, pp.130-145. [2] M. Prensky (2001) Digital Game-Based Learning. McGraw-Hill. [3] M.M. Klawe (1998) When Does The Use Of Computer Games And Other Interactive Multimedia Software Help Students Learn Mathematics? http://www.cs.ubc.ca/nest/egems/reports/NCTM.doc [4] K. Maragos, M. Grigoriadou (2005) Towards the design of Intelligent Educational Gaming Systems. Proc. of AIED05 WORKSHOP5, pp.35-38. [5] H.M. HALFF (2005) Adventure Games for Science Education: Generative Methods in Exploratory Environments. Proc. of AIED05 WORKSHOP5, pp.12-20. [6] T. Umetsu, T. Hirashima, A. Takeuchi (2004) Partial Exchange Method for Designing Learning Games and Its Application. Proc. of ICCE2004, pp. 257-264.