In this paper, we present a uni ed model for creativitytechnique based processes that considers the key properties of each of these creativity techniques. For the construction of this model we rst analyze processes of various creativity techniques with respect to their key properties. Afterwards, we use these ndings to formalize a uni ed model and discuss its use for more exible creativity support systems.
Creativity techniques are used in many domains to guide creative problem
solving processes [
Collaborative tools for the support of creativity-technique based problem
solving processes should address the main shortcomings of theses processes which
may also include typical problems for interacting groups such as the factors
production blocking, group pressure and social loa ng [
As of today, CSS are tailored for a speci c creativity technique (e.g. Mindmanager1) or have a quite limited portfolio of supported techniques for idea generation (divergent phase) and idea evaluation (convergent phase) (e.g. Thinktank2). Hence, for a given context (i.e. problem type, group, situation, etc.) the available techniques may be not appropriate or be less e ective than an alternative technique. Furthermore, using di erent techniques on the same problem can be bene cial for the process.
In order to support a broad set of creativity techniques, CSS need a unied model of creative processes that considers each of these techniques. As in any modeling process, the question of the appropriate level of abstraction has to be faced: A higher level of abstraction increases exibility, but decreases the semantic information. E.g. a collaborative drawing tool on a virtual whiteboard may be very exible for the use of creativity techniques. However, because of 1 http://www.mindjet.com 2 http://www.groupsystems.com/technology/thinktank its high level of abstraction, the key properties of creativity-technique based processes are not comprehensively supported. This includes e.g. the possibility to anonymize group members or to set time limits for the divergent phase as required in speci c techniques. Additionally the semantic of the single user contributions may be lost (e.g. no di erentiation between ideas, evaluations or messages in the system).
In this paper, we present a uni ed model for creativity-technique based processes that considers the key properties of each of these creativity techniques. For the construction of this model we rst analyze processes of various creativity techniques with respect to their key properties. Afterwards, we use these ndings to formalize a uni ed model. This model can in turn be used to implement a corresponding software system that can support di erent creative-technique based problem solving processes.
The aim of the process analysis is to examine creative problem solving processes and to identify the key properties that can have a positive impact on the process outcomes. Obviously, a process model of a CSS should focus on these properties.
Creative processes are typically seen as a sequence of divergent and
convergent process phases [
The abstract perspective on the process is helpful to set the frame for a model
for creative processes, but no advice is given on how the activities within the
phases can be supported. Since creativity techniques claim to support creative
problem solving processes, they can be regarded as a source for such
parameters. Given their highly practical orientation, it is likely that they can give more
concrete advice than an abstract process model. Based on this idea, we analyzed
a multitude of common creativity techniques [
In the meanwhile, we examined 14 additional creativity techniques3 and
further re ned the list of important process phase parameters in
creativitytechnique based processes. The following parameters were found for both phase
types:
3 For descriptions of all mentioned creativity techniques in this article see [
The following parameters were found for the divergent phases only: { Stimuli: Mental stimuli are pieces of information that are presented to participants in order to in uence their thinking processes during the convergent phases. Stimuli are only rarely statically de ned by the creativity technique itself. Instead, they are often contributed by the participants in previous phases (e.g. random stimuli technique). { Start ideas: Start ideas are available from the beginning of a divergent phase and can directly be used to combine new ideas with. The ideas are typically generated in preceding divergent phases (e.g. morphological analysis ). { Constraints of idea representations: While for most techniques a verbal or textual representation of the generated ideas is assumed, a few techniques restrict the way the participants may express their ideas. The greeting card technique speci es the use of pictures that should be combined to compose ideas. During brainsketching, participants can only use sketches to explain their ideas. { Limitation of idea quantity: Looking futile at the rst glance, bounding the idea quantity may be necessary for practical reasons (e.g. Brainwriting-6-35 ).
The following parameters were found for the convergent phases only: { Scenarios: A scenario is the description of a plausible future. A scenario de nes a hypothetical context for the idea evaluation. The scenarios may be de ned by the technique itself (e.g. four future technique) or may be generated by the participants in a previous process. { Criteria: With criteria, the evaluation of ideas can be restricted to a certain aspect. Techniques mostly de ne static criteria (e.g. the castle technique de nes e ectivity, practicality and originality), but the criteria could also be dynamically generated by the participants in a previous process. { Scoring: Scoring refers to assigning numerical values to ideas. Many techniques for convergent process phases imply using scores of a given range as evaluation measure (e.g. sticking dots technique). Idea evaluations from numerous participants expressed as scores can be easily processed and merged (e.g. by computing average values). { Comments: By using comments, participants can evaluate ideas with free texts. With comments, participants can advance their opinion in a much more detailed way than with scoring. However, it is harder to aggregate these pieces of information than in the case of scoring.
The parameters presented in the previous section can be regarded as requirements for a uni ed process model since we empirically found out that the processes implied by the techniques can be seen as a sequence of divergent and convergent phases based on these parameters. For being used in the context of a computer system, the process model has to be formalized. For the process phases, we suggest the formalization depicted in gure 1: ProcessPhase acts as
ProcessPhase problem : Problem participants: Set<Person> ideas : Set<Idea> timelimit : Integer anonymous : Boolean
DivergentPhase stimuli : Set<Stimulus> maxIdeas : Integer allowSketch : Boolean allowText : Boolean allowImage : Boolean
ConvergentPhase scenario : Scenario criterion : Criterion allowComments : Boolean allowScoring : Boolean maxScore : Integer the base class for the two di erent types of process phases. ProcessPhase must declare the following attributes: { problem: The problem that should be solved in the process phase. Problems are commonly represented as plain text strings, but as concrete implementations of CSS (e.g. for special domains) may have additional requirements we suggest to de ne a dedicated class Problem. { participants: The persons that are participating in the process phase. { ideas: The ideas of the process phase. For divergent phases this attribute can be initialized with start ideas if needed, but in most cases the set is empty at the beginning of a divergent phase. In convergent phases, the set is initialized with the ideas from the precedent divergent phase so the ideas can be evaluated by the users. The results of the process analysis makes some implications about the functionality of the Idea class. For the divergent phases, it should support the expression of ideas as text, with images and with sketches. For the convergent phases, it should be able to store user ratings as numerical values and as text. Further adoption to a target domain of a CSS can be necessary.
The remaining attributes of ProcessPhase are formalizations of the creativity technique parameters found for both types of process phases: { timelimit: Timelimits in a process phase are represented with an integer value that stores the remaining seconds for the phase or is unde ned if no timelimit is set. { anonymous: The need of anonymity can be represented by a boolean ag.
If the ag is set, the CSS must keep all person-related information hidden during the process phase.
The divergent phases are represented by the class DivergentPhase, having the following attributes in addition to the ones from the base class: { stimuli: A set of mental stimuli which must be presented to the participants by the CSS. If the value is not set, no stimuli are used in this divergent phase. Since stimuli are often generated by the group in previous phases and thus are formulated as ideas, the Stimuli class should be modeled similar to the Idea class or the same classes should be used. { maxIdeas: Integer value that limits the number of ideas that may be generated in the process phase. If the value is not set, the number of ideas is not limited. { allowSketch: If the boolean value is set, the users may sketch to express their ideas. { allowText: If the boolean value is set, the users may use text to express their ideas. { allowImage: If the boolean value is set, the users may use images to express their ideas.
The convergent phases are represented by the class ConvergentPhase, having the following attributes in addition to the ones from the base class: { scenario: The scenario that should be considered when evaluating the ideas in the process phase. In most cases it will be su cient to represent scenarios as a string describing the particular scenario, but a dedicated class is preferable. If no scenario should be considered, the value is not set. { criterion: The criterion that should be evaluated in the process phase. For the representation of criteria, the same considerations apply as for scenarios.
If no criterion is set, the idea is to be evaluated as a whole. { allowComments: If the boolean value is set, the users may evaluate the ideas using textual comments. { allowScoring: If the boolean value is set, the users may assign scores to the ideas. { maxScore: If scoring is allowed, this value de nes the scoring range (from 0...maxScore).
The simplicity of the presented formalization for process phases of creativitytechnique based processes makes it easy to be used in CSS implementations, yet it is powerful enough to represent creative problem solving processes of numerous creativity techniques and even combinations, since it contains all the key properties found in our process analysis.
To illustrate the use of the uni ed process model, we present formalizations of creative processes that correspond to some well-known creativity techniques. As explained beforehand, a creative process is understood as a sequence of divergent and / or convergent phases, each having a particular set of attribute values. The complete set of possible attributes was presented in the previous section. In this example section, for clarity we will omit attributes that are not important for the given creativity technique. For the not-listed attributes, default values (e.g. "no restriction" for restricting attributes or "arbitrary number of participants" for the participant attribute) can be assumed.
The major principle of the brainstorming technique is to avoid any idea evaluation during the idea generation
Begin phase. Since idea evaluation is not possible in divergent DivergentPhase phases of the proposed model, this principle is achieved problem: "How could the problem by modeling brainstorming as a divergent phase. Addii#dPeabseasroelvweedl?coWmiled. aTnrydtuonbuusiuldalon tional information to the given problem (here labeled tahneoindyemasouosf:ottrhueer participants." #P) in the problem attribute makes the participants aware of the remaining brainstorming principles (wild End ideas, building up on ideas of others). The classic brainstorming needs no further attribute values in the divergent phase, since no further restrictions are made by the technique. Alternative brainstorming variants can be achieved by setting attributes of the phase, e.g. Anonymous Brainstorming by setting the anonymous attribute (as depicted in the gure on the left) or Brainsketching by setting allowSketch to true and allowText as well as allowImage to false, so only sketches can be used for expressing ideas.
Begin
DivergentPhase problem: "Find three solutions for the problem #P." participants: U1 timelimit: 300 maxIdeas: 3
DivergentPhase problem: "Find three solutions for the problem #P." participants: U2 timelimit: 300 maxIdeas: 3 {I4,I5,I6} {I1,I2,I3} problem: D"TivryertogeimntpProhvaesethese three ideas of your group member." participants: U1 timelimit: 300 ideas: I1,I2,I3
End problem: D"TivryertogeimntpProhvaesethese three ideas of your group member." participants: U2 timelimit: 300 ideas: I4,I5,I6
A creative process with two participants U1, U2 based on the Brainwriting 6-3-5 technique can be modeled by DivergentPhase objects as shown in the gure to the left. In the rst round (upper two phases), the participants are asked to nd three solutions for the given problem. The participants have to work separately on their solutions, so U1 and U2 are in separate phases (participants attribute). The technique imposes a timelimit of 5 minutes, which is modeled with a value of 300 for the timelimit attribute, and sets an upper limit of 3 ideas with the maxIdeas attribute. When the timelimit has exceeded, the ideas generated by the participants are exchanged and placed in the ideas attributes of the successive phases. The participants are now asked to improve the received ideas instead of generating completely new ones.
Begin
DivergentPhase problem: "How could you improve an existing solution for your problem #P by substituting place, time, material or people?"
DivergentPhase problem: "What materials, features, processes, people, products etc. could you combine to find better solutions for your problem #P?"
DivergentPhase problem: "How could you improve an existing solution for #P by altering or changing the function of elements?" ...
End
Begin
ConvergentPhase criterion : "Effectivity" allowComments : false allowScoring : true maxScore : 1
ConvergentPhase criterion : "Practicality" allowComments : false allowScoring : true maxScore : 1
ConvergentPhase criterion : "Originality" allowComments : false allowScoring : true maxScore : 1
End
The SCAMPER technique can be modeled as a sequence of divergent phases, where in each phase a slightly di erent approach towards a solution is suggested using the problem attribute of the phase object. The gure on the left shows the rst 3 phases of the SCAMPER technique (Substitute, Combine, Adapt), the remaining phases (Modify, Put to another use, Eliminate and Reverse) are modeled similarly. This way, all type of checklist-based technique (e.g. Osborn-checklist, CATWOE) can be represented. Since the techniques impose no further restrictions, the remaining phase attributes are not set.
By setting some of the attributes, combinations of techniques could be achieved: e.g. setting the anonymous attribute in the phases leads to an Anonymous SCAMPER process.
The Castle Technique is an evaluation technique that suggests to judge ideas in sense of their e ectivity, practicality and originality. To speed up the decision process, the participants may only say if the criterion is met or not. Formalized by the uni ed process model, a castle technique process is a sequence of convergent phases (ConvergentPhase objects). In each phase, the participants have to decide if the given ideas (in general coming from a previous divergent phase) ful ll the criterion de ned by the attribute criterion. To express their decision, participants may only use score values (allowScoring is true while allowComment is false) and can only make a binary decision, since the maxScore attribute is set to 1. In this article we proposed a formalization for creativity-technique based problem solving processes as sequence of process phases. We rst described how creativity techniques are currently supported in creative support systems. Then, we argued why a uni ed process model is a way towards more exible CSS. In the second section we summarized the results of our analysis of a large set of creativitytechniques with respect to their key properties. Finally, we presented a formal uni ed model comprising the key properties of creativity-technique based problem solving processes and illustrated the approach with some example process formalizations.
Since creativity techniques guide creative processes by a ecting the parameters we identi ed in the process analysis, they can be interpreted as presets of attribute values in the process model (e.g. a speci c con guration of the attributes timelimit, stimuli etc.). Following this concept, a formalization of creativity techniques (in contrast to concrete creative processes) can be achieved.
Furthermore, the key properties of creativity techniques can be regarded as a framework for analyzing creativity techniques themselves: by varying single attribute values of the speci c process phases it is possible to investigate the e ects on the produced outputs. For instance by varying the timelimit attribute value in di erent creative processes, a better general understanding on the e ect of time limits in creative problem solving processes can be gained.
As a framework for evaluation, we are currently implementing a CSS prototype which is based on the here discussed uni ed process model. Thereby, our goal is to assess the completeness of the proposed key components.