Collaborative Ideation Partner: Design Ideation in Human-AI

Co-creativity

Jingoog Kim

, Mary Lou Maher and Safat Siddiqui

University of North Carolina at Charlotte, 9201 University City Blvd., Charlotte, NC 28223, U.S.A.

Keywords: Co-creativity, AI-based Co-creative System, Ideation, Design.

Abstract: AI-based co-creative design systems enable users to collaborate with an AI agent on open-ended creative

tasks during the design process. This paper describes a co-creative system that supports design creativity by

providing inspiring design solutions in the initial idea generation process, based on the visual and conceptual

similarity to sketches drawn by a designer. The interactive experience allows the user to seek inspiration

collaborating with the AI agent as needed. In this paper, we study how the visual and conceptual similarity of

the inspiring design from the AI partner influences design ideation by examining the effect on design ideation

during a design task. Our findings show that the AI-based stimuli produce ideation outcomes with more

variety and novelty when compared to random stimuli.

1 INTRODUCTION

Computational co-creative systems are a growing

research area in computational creativity. While some

research on computational creativity has a focus on

generative creativity (Colton et al., 2012; Gatys et al.,

2015; Veale, 2014), co-creative systems focus on

computer programs collaborating with humans on a

creative task (N. M. Davis, 2013; Hoffman &

Weinberg, 2010; Jacob et al., 2013). Co-creative

systems have enormous potential since they can be

applied to a variety of domains associated with

creativity and encourage designers’ creative thinking.

Understanding the effect of co-creative systems in the

ideation process can aid in the design of co-creative

systems and evaluation of the effectiveness of co-

creative systems. However, most research on co-

creative systems focuses on evaluating the usability

and the interactive experience (Karimi et al., 2018)

rather than how the co-creative systems influence

creativity in the creative process. In this paper we

focus on ideation rather than the user experience in

order to understand the cognitive effect of AI

inspiration.

Ideation, an idea generation process for

conceptualizing a design solution, is a key step that

can lead a designer to an innovative design solution

https://orcid.org/0000-0003-3597-6153

in the design process. Idea generation is a process that

allows designers to explore many different areas of

the design solution space (Shah et al., 2003). Ideation

has been studied in human design tasks and

collaborative tasks in which all participants are

human. Collaborative ideation can help people

generate more creative ideas by exposing them to

ideas different from their own (Chan et al., 2017).

Recently, the field of computational creativity began

exploring how AI agents can collaborate with humans

in a creative process. We posit that a co-creative

system can augment the creative process through

human-AI collaborative ideation.

We present a co-creative sketching AI partner, the

Collaborative Ideation Partner (CIP), that provides

inspirational sketches based on the visual and

conceptual similarity to sketches drawn by a designer.

To generate an inspiring sketch, the AI model of CIP

computes the visual similarity based on the vector

representations of visual features of the sketches and

the conceptual similarity based on the category names

of the sketches using two pre-trained word2vec

models. The turn-taking interaction between the user

and the AI partner is designed to facilitate

communication for design ideation. The CIP was

developed to support an exploratory study that

evaluates the effect of an AI model for visual and

Kim, J., Maher, M. and Siddiqui, S.

Collaborative Ideation Partner: Design Ideation in Human-AI Co-creativity.

DOI: 10.5220/0010640800003060

In Proceedings of the 5th International Conference on Computer-Human Interaction Research and Applications (CHIRA 2021), pages 123-130

ISBN: 978-989-758-538-8; ISSN: 2184-3244

123

conceptual similarity on design ideation in a co-

creative design tool.

In this paper, we emphasize the effect of the AI-

based inspirations based on the visual and conceptual

similarity. The main contributions of this paper to the

HCI and co-creativity community are (1) a

methodology for evaluating the impact of AI

inspiration on ideation and (2) the impact of AI-based

visually and conceptually similar designs on ideation.

2 COMPUTATIONAL

CO-CREATIVE SYSTEMS

Computational co-creative systems are one of the

growing fields in computational creativity that

involves human users collaborating with an AI agent

to make creative artifacts. The distinction of co-

creativity from computational creativity is that co-

creativity is a collaboration in which multiple parties

contribute to the creative process in a blended manner

(Mamykina et al., 2002). Co-creative systems have

been applied in different creative domains such as art,

music, dance, drawing, and game design. Some co-

creative systems directly perform actions on a shared

artifact or contribute to a performance whereas others

provide suggestions to inspire users for generating

novel ideas. This distinguishes how a co-creative AI

agent contributes to the creative process. One co-

creative interaction paradigm is an AI agent

performing actions with a user simultaneously.

Shimon (Hoffman & Weinberg, 2010) is a robotic

marimba player that listens and responds to a

musician in real time. This improvisational robotic

musician performs accompaniment with the users’

musical performance simultaneously. Another co-

creative interaction paradigm is a turn-taking action

between a user and an AI agent in a shared artifact.

Drawing Apprentice (N. Davis et al., 2015) is a co-

creative drawing system in which the computational

partner analyzes the user's sketch and responds to the

user’s sketch. Viewpoints AI (VAI) is a co-creative

dance partner that analyzes the user’s dance gestures

and provides complimentary dance in real-time by a

virtual character projected on a large display screen

(Jacob et al., 2013). These co-creative interaction

paradigms are examples of an AI agent participating

in a creative activity by performing the same type of

action as a user. Another co-creative interaction

paradigm is providing suggestions to the user.

Sentient Sketchbook (Yannakakis et al., 2014) and

3Buddy (Lucas & Martinho, 2017) are co-creative

systems for game level design. In both systems, the

AI agent provides feedback and additional ideas to

develop the game design rather than creating game

level directly.

3 THE COLLABORATIVE

IDEATION PARTNER (CIP)

The Collaborative Ideation Partner (CIP) as a co-

creative design system builds on previous projects

(Karimi et al., 2019, 2020) that interpret sketches

drawn by a user and provides inspirational sketches

based on visual similarity and conceptual similarity.

We developed the CIP to explore the effect of an AI

model for visual and conceptual similarity on design

ideation in a co-creative design tool.

The user interface of CIP is shown in Figure 1.

There are two main spaces in the CIP interface: the

drawing space (pink area) and the inspiring sketch

space (purple area). The drawing space consists of a

design task statement, undo button, clear button, and

user’s canvas. The design task statement in the

drawing space includes the object to be designed as

well as a context to further specify the objects’ use

and environment. The user can draw a sketch in the

drawing space and edit the sketch using the undo and

clear button. The inspiring sketch space includes an

“inspire me” button, the name of the inspiring object,

and a space for presenting the AI partner’s sketch.

When the user clicks the “inspire me” button after

sketching their design concept, the AI partner

provides an inspiring sketch based on visual and

conceptual similarity. An ideation process using CIP

involves turn-taking communications between the

user and the AI partner. Another part of the CIP

interface in addition to the two main spaces is the top

area (grey area) including a hamburger menu and an

introductory statement. The hamburger menu on the

top-left corner of the interface includes four design

tasks (i.e. sink, bed, table, chair) and allows the

experiment facilitator to select one of the design

tasks. Each design task provides different categories

of ideation stimuli.

Figure 1 shows an example of an inspiring sketch

and how participants communicate with an inspiring

sketch to develop their design. The design task shown

in Figure 1 is to design a chair for a gaming computer

desk. The participant drew a basic chair with back,

seat, legs, and small wheels before requesting

inspiration from the AI partner. The sketch suggested

from the AI models is a bulldozer: visually similar

and conceptually different to the participant’s sketch.

After getting the inspiring sketch, the participant

CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications

124

Figure 1: User interface of Collaborative Ideation Partner.

made the wheels much bigger for better mobility and

added a leg rest for comfort. During the retrospective

protocol, the participant described that “I decided to

go with bigger wheels here, just thinking of bulldozer,

little more heavy duty. I mean, I also noticed the little

lift gate or whatever that is. And that kind of made me

think that I needed to add like some kind of leg

support and that kind of made sense.”

3.1 Dataset

For the source of inspiring sketches, CIP uses a public

benchmark dataset called QuickDraw! (Jongejan et

al., 2016), which was created during an online game

where players were asked to draw a particular object

within 20 seconds. The dataset includes 345

categories with more than 50 million labelled

sketches, where sketches are the array of the x and y

coordinates of the strokes. The system uses the

simplified drawing json files that use Ramer–

Douglas–Peucker algorithm (Douglas & Peucker,

1973; Ramer, 1972) to simplify the strokes, and

position and scale the sketches into a 256 X 256

region. The stroke data associated with these sketches

are used to calculate the visual similarity and the

corresponding category names are used to measure

the conceptual similarity.

3.2 AI Models for Visual and

Conceptual Similarity

The CIP has 2 distinct components for measuring

similarity between the user’s sketch and the sketches

in the dataset: one component for calculating visual

similarity and another component for calculating

conceptual similarity. The visual similarity

component selects sketches from the sketch dataset

based on a representation of the stroke data in the

image file. The conceptual similarity component

computes the degree of similarity between the

category names of the objects in design tasks and the

category names in the objects in the sketch dataset.

For the visual similarity component, we used a

pre-trained CNN-LSTM model from the precedent

with 3 convolutional layers, 2 LSTM layers, and a

softmax output layer on the QuickDraw dataset

(Karimi et al., 2019, 2020). For the conceptual

similarity component, we considered sketch category

names in the QuickDraw dataset as the concepts of

the sketches that contain 345 unique categories. We

used two pre-trained word2vec models, Google News

(Mikolov et al., 2013) and Wikipedia (Rehurek &

Sojka, 2010), and calculated cosine similarities for

measuring the conceptual similarities between the

object categories of the design tasks and the

categories of inspiring sketches from the dataset. For

each category of the design tasks, we generated two

sorted lists of conceptually similar category names,

one for each word2vec model, and then used human

judgement to compare the sorted lists and select the

top 15 common conceptually similar category names

that appear in both lists. This final step of using

human judgement improved the alignment between

the conceptual similarities of AI models and human

perception. The conceptual similarity component of

CIP uses the common list of category names for

sorting the sketches based on the conceptual

similarities.

We use these two AI-based components of the

CIP to generate sequences of sketches with

combinations of visual and conceptual similarity to

Collaborative Ideation Partner: Design Ideation in Human-AI Co-creativity

125

the user’s current sketch and design task to inspire the

user during their design process and measure the

effect of visual and conceptual similarities on

ideation.

3.3 AI-based Inspiration in CIP

To support an exploratory study that measures

ideation when co-creating with CIP, the interaction

with CIP has four distinct modes of inspiration that

vary the visual and conceptual similarity. Each of the

four modes appears as a design task (i.e. sink, bed,

table, chair) in the CIP interface. One of the modes

(i.e. sink) uses a random sketch selection while three

other modes use AI models to select an inspiring

sketch as inspiration in CIP.

 Random: Inspire with a random sketch (sink):

The CIP selects a sketch randomly from the

sketch dataset to be displayed on the AI

partner’s canvas.

 Similar: Inspire with a visually and

conceptually similar sketch (bed): The CIP

selects a sketch from a set of sketches where

each one is similar visually and conceptually to

the user’s sketch (e.g. user sketch - a bed, AI

sketch - a similar shape of bed to the user’s

sketch).

 Conceptually Similar: Inspire with a

conceptually similar and visually different

sketch (table): The CIP selects a sketch from a

set of sketches where each one is conceptually

similar but visually different to the user’s

sketch (e.g. user sketch - a square table, AI

sketch - a round table).

 Visually Similar: Inspire with a visually

similar and conceptually different sketch

(chair): The CIP selects a sketch from a set of

sketches where each one is visually similar but

conceptually different to the user’s sketch (e.g.

user sketch - a circular chair back, AI sketch -

a face).

4 EXPLORATORY STUDY

The goal of the exploratory study is to explore the

effect of AI inspiration on ideation through an

analysis of the correlation between conceptual and

visual similarity with characteristics of ideation.

Specifically, we are interested in the relationship

between the users’ ideation and sources of AI

inspiration.

4.1 Study Design

The type of study is a mixed design of between-

subject and within-subject design. There are 3 groups

of within-subject design (i.e. A&B, A&C, A&D) in

this study and each group has a control condition (i.e.

condition A) and one of 3 treatment conditions (i.e.

condition B, C, D). The control condition (condition

A) for each group is the same but the treatment

condition for each group is different (condition B or

C or D). The control condition and 3 treatment

conditions are the different types of inspirations

presented in Section 3.3:

 Condition A (control condition): randomly

(sink)

 Condition B (treatment condition): visually and

conceptually similar (bed)

 Condition C (treatment condition):

conceptually similar and visually different

(table)

 Condition D (treatment condition): visually

similar and conceptually different (chair)

The protocol including the informed consent

document has been reviewed and approved by our

IRB and we obtained informed consent from all

participants to conduct the experiment. We recruited

12 students from human-centered design courses for

the participants: each participant engaged in 2

conditions: a control condition and one of the

treatment conditions, with 4 participants for each of

the 3 groups of within-subject design (i.e. A&B,

A&C, A&D). The experiment is a mixed design with

N=4 and a total of 12 participants.

The task is an open-end design task in which

participants were asked to design an object in a given

context through sketching. Different objects for the

design task were used for each condition: a sink for

an accessible bathroom (condition A), a bed for a

senior living facility (condition B), a table for a

tinkering studio, a collaborative space for designing,

making, building, etc. (condition C), a chair for a

gaming computer desk (condition D).

The procedure consists of a training session, two

design task sessions, and two retrospective protocol

sessions. In the training session, the participants are

given an introduction to the features of the CIP

interface and how they work to enable the AI partner

to provide inspiration during their design task. After

the training session, the participants perform two

design tasks in a control condition and a treatment

condition. The study used a counterbalanced order for

the two design tasks. The participants were given as

CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications

126

much time as needed to perform the design task until

they were satisfied with their design. The participants

are free to click the “inspire me” button as many times

as they would like to get inspiration from the system.

However, the participants were told to have at least 3

inspirational sketches (i.e. clicking the “inspire me”

button at least 3 times during a design session), a

minimum number of inspirations, from the system.

Once the participants finish the two design task

sessions, the participants are asked to explain what

they were thinking while watching their design

session recording as time goes on, and how the AI's

sketches inspired their design in the retrospective

protocol session.

4.2 Data Collected

Two types of data were collected for analyzing the

study results: a set of sketches that participants

produced during the design tasks and the

verbalization of the ideation process during the

retrospective protocol. We recorded the entire design

task sessions and retrospective sessions for each

participant. The sketch data collected from the

recordings of design task sessions shows the progress

of design and the final design visually for each design

task session. The verbal data collected from the

recordings of retrospective sessions records how the

participants came up with ideas collaborating with the

AI partner and applied the ideas to their design.

4.3 Data Segmentation and Coding

To analyze the verbal data collected from the

retrospective sessions, we adapted the FBS coding

scheme for characterizing cognitive issues during a

design process (Gero, 1990; Gero & Kannengiesser,

2004). An idea can be variously defined as a

contribution that contains task-related information, a

solution in the form of a verb–object combination,

and a specific benefit or difficulty related to the task

(Reinig et al., 2007). The FBS coding scheme

provides a segmentation into individual ideas

associated with specific cognitive issues in design.

First, the verbal data of all retrospective protocol

sessions was transcribed. The transcripts were

segmented based on the inspiring sketches the

participant clicked. A segment starts with an inspiring

sketch and ends when the inspiration is clicked for the

next sketch. To identify each idea in an inspiring

sketch segment, we segmented the inspiring segments

again based on FBS ontology (Gero, 1990; Gero &

Kannengiesser, 2004) as an idea segment, since an

inspiring sketch segment includes multiple ideas. The

idea segments were coded based on FBS ontology

(Gero, 1990; Gero & Kannengiesser, 2004) as

requirement (R), function (F), expected behavior

(Be), behavior from structure (Bs), and structure (S).

A segment coded R is an utterance that talks about the

given requirement in the statement of design task (e.g.

accessible bathroom); a segment coded F is an

utterance that talks about a purpose or a function of

the design object (e.g. more accessible); a segment

coded Be is an utterance that talks about an expected

behaviors from the structure (e.g. water could

automatically come out); a segment coded Bs is an

utterance that talks about a behavior derived from the

structure (e.g. pressing on); a segment coded S is an

utterance that talks about a component of the design

object (e.g. button). The result of this coding scheme

is a segmentation of the verbal protocol into

individual ideas, each associated with one code: R, F,

Be, Bs, S.

Two coders coded the idea segments individually

based on the coding scheme above then came to

consensus for the different coding results. The coding

instruction was given to the coders included how to

segment inspiring sketch segments and idea

segments, how to code each idea segment with the

coding scheme, and how to code new and repeated

ideas. The two coders coded a design session together

to make an initial agreement for segmentation and

coding before coding individually then coded all

design sessions individually. Once each coder

completed coding all data individually, the two

coders discussed each of the different coding results

and came to consensus.

4.4 Analysis of Exploratory Study:

Measuring Ideation

To evaluate the effect of AI inspiration on ideation,

we adapted the metrics from Shah et al. (2003) for

measuring ideas in a design process. We applied four

types of metrics for measuring ideation effectiveness,

used for evaluating idea generation in design:

novelty, variety, quality, and quantity of design ideas.

We developed the four metrics based on (Shah et al.,

2003) to analyze the coded data of the retrospective

protocol session.

Novelty. Novelty is a measure of how unusual or

unexpected an idea is as compared to other ideas

(Shah et al., 2003). In this study, a novel idea is

defined as a unique idea across all design sessions in

a condition. For measuring novelty, we counted how

many novel ideas in the entire collection of ideas in a

Collaborative Ideation Partner: Design Ideation in Human-AI Co-creativity

127

Figure 2: The number of novel ideas in the group of A&C.

design session (personal level of novelty) and a

condition (condition level of novelty). We removed

the same ideas across all design sessions in a

condition then counted the number of ideas.

The results showed that all treatment conditions

(B, C, D) have more novel ideas than the control

condition (A) in the total number of novel ideas.

Specifically, 10 participants out of 12 participants

produced more novel ideas in a treatment condition

than the control condition. When comparing the

novelty of 3 groups, the group A&C showed the

largest difference between the control condition and

the treatment condition where condition C selected

inspiring sketches that are conceptually similar and

visually different. As shown Figure 2, all participants

in the group of A&C produced more novel ideas in

the condition C than the condition A while one of the

participants (i.e. P4) in the group A&B and one of

participants (i.e. P9) in the group A&D produced

fewer novel ideas in the treatment condition than the

control condition. This result can indicate that the

conceptual similarity of inspiring sketches may be

associated with the novelty of ideas in the ideation

with CIP.

Variety. Variety is a measure of the explored solution

space during the idea generation process (Shah et al.,

2003). Each idea segment was coded whether it is a

new idea or a repeated idea in a design session. For

measuring variety in this study, only the number of

new ideas coded as R/F/B/S is counted in a design

session while the metric of quantity includes both

new ideas and repeated ideas.

The results showed that the variety of ideas in

condition C is higher than in condition A. Figure 3

shows the results of codes comparing the control

condition (A) and one of the treatment conditions (C).

The results of the group A&C show some distinct

patterns in function. All participants produced more

functions in condition C than in condition A. The

number of function ideas showed a large difference

Figure 3: Variety of ideas in the group of A&C.

for all participants between condition A and C. This

result indicates that the conceptual similarity inspired

the participants to produce more various functions

associated with the context of the design.

Quality. Quality is a subjective measure of the design

(Shah et al., 2003). In this study, quality is measured

using the Consensual Assessment Technique (CAT)

(Amabile, 1982), a method in which a panel of expert

judges is asked to rate the creativity of projects. Two

judges, researchers involved in this study,

individually evaluated the final design in each

condition as low/medium/high quality, in two

evaluation rounds. In the first-round of evaluation,

each judge evaluated the final designs identifying

some criteria for evaluating the quality of ideas. Once

the judges finished the first-round of evaluation, they

shared the criteria they identified/used, not sharing

the results of the evaluation, then made a consensus

for the criteria that will be used for the second-round

evaluation. The criteria that the judges agreed for

evaluating the quality of ideas in this study are the

number of features, how responsive the features are

to the specific task, how creative the design is. In the

second-round evaluation, each judge evaluated the

final design again using the agreed criteria.

Table 1: Quality evaluation results of each judge in the

group of A&D.

Condition A Condition D

Judge 1 Judge 2 Judge 1 Judge 2

P3 low low hi

h Hi

P5 low low mediu

mediu

P9 mediu

mediu

h Hi

P12 low low low Low

The results showed that the quality of ideas in

condition D is higher than in condition A, where

condition D selects sketches that are visually similar

P2 P8 P6 P11

The number of ideas

Condition A Condition C

The number of ideas

R F B S

CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications

128

and conceptually different for inspiration. Table 1

shows the result of the quality evaluation that each

judge made for each design in condition A and

condition D. Three out of four participants produced

higher quality in condition D than condition A. P3

produced much higher quality in condition D than

condition A (i.e. low to high). P5 and P9 produced

higher quality in condition D than in condition A (i.e.

P5: low to medium, P9: medium to high). This result

indicates that the visual similarity of inspiring

sketches may be associated with the quality of ideas

in the ideation with CIP.

Quantity. Quantity is the total number of ideas

generated (Shah et al., 2003). For measuring quantity

in this study, the number of ideas both new ideas and

repeated ideas coded as R/F/B/S is counted in a

design.

Figure 4: Quantity of ideas in the group of A&C.

Figure 4 shows the results of the quantity of ideas

in the group of A&C. The results show a similar

pattern to the result of variety with some distinct

patterns. First, for the total number of ideas, 3 out of

4 participants (i.e. P2, P8, P6) generated more ideas

in condition C than in condition A. Second, 3 out of 4

participants (i.e. P2, P8, P6) generated more ideas of

F (function) and S (structure) in condition C than in

condition A. This result indicates that the conceptual

similarity of inspiring sketches facilitates producing

new functions and the emerging functions were

transferred to structures of the design.

Our exploratory study does not have a sufficient

number of participants to allow us to generalize the

results for all cases of ideation from AI-based visual

and conceptual similarity. However, we did a

significance test on the results to see if there are

significant trends to look for in a more robust study.

A paired t-test was conducted to determine the

significance of our results between the control

condition and the treatment conditions in novelty,

variety, and quantity. The results showed a significant

difference in variety and quantity. For variety,

participants in condition C (M=24.25, SD=10.21)

produced more functions than in condition A

(M=15.50, SD=9.81), t(3)=−5.14, two tail

p=0.014253. For quantity, participants in condition B

(M=28.75, SD=12.76) produced more functions than

in condition A (M=19.00, SD=13.24), t(3)=−3.30,

two tail p=0.045732. This exploratory study does not

have enough participants to measure or check for

statistical significance, but the trends of the results

show the potential for further analysis of the effect of

an AI model for visual and conceptual similarity on

design ideation with the metrics we identified for

measuring ideation.

5 DISCUSSION

In this paper, we presented a co-creative design

system called CIP and an exploratory study that

explores the effect of an AI model for visual and

conceptual similarity on design ideation in a co-

creative design tool. To evaluate the effect of AI

inspiration on ideation, we applied four metrics (i.e.

novelty, variety, quality, quantity) to measure the

ideation in an exploratory study. Overall our findings

show that the AI-based stimuli produce different

ideation outcomes when compared to random stimuli.

More specifically, we found that different types of

AI-based stimuli show potential for different types of

ideation. Novel ideation is associated with AI-based

conceptually similar stimuli. Idea variety and quantity

is associated with both AI-based visual and

conceptual similarity of the inspiration. Idea quality

is associated with visual similarity.

In addition to measuring ideation, we observed the

video stream data to see how participants develop

their design ideas communicating with the

inspirations. The participants' responses to

inspirations showed different patterns of users on the

use of CIP in an ideation process. In an evolution of

the participant’s sketch, participants in each condition

start with a basic shape of the target design then

develop the design with inspiration from the AI

partner. Participants explored many inspiring

sketches in condition A but did not have many design

changes; while participants in conditions B, C, and D

developed their design in response to fewer inspiring

sketches. This observation suggests further analysis

of ideation to understand the cognitive process of

ideation when co-creating with the CIP.

100

The number of ideas

R F B S

Collaborative Ideation Partner: Design Ideation in Human-AI Co-creativity

129

6 CONCLUSION

This paper presents a co-creative design tool called

Collaborative Ideation Partner (CIP) that supports

idea generation for new designs with stimuli that vary

in similarity to the user’s design in two dimensions:

conceptual and visual similarity. The AI models for

measuring similarity in the CIP use deep learning

models as a latent space representation and similarity

metrics for comparison to the user’s sketch or design

concept. The interactive experience allows the user to

seek inspiration when desired. To study the impact of

varying levels of visual and conceptual similar

stimuli, we performed an exploratory study with four

conditions for the AI inspiration: random, high visual

and conceptual similarity, high conceptual similarity

with low visual similarity, and high visual similarity

with low conceptual similarity. To evaluate the effect

of AI inspiration, we evaluated the ideation with CIP

using the metrics of novelty, variety, quality and

quantity of ideas. We found that conceptually similar

inspiration that does not have strong visual similarity

leads to more novelty, variety, and quantity during

ideation. We found that visually similar inspiration

that does not have strong conceptual similarity leads

to more quality ideas during ideation. Future AI-

based co-creativity can be more intentional by

contributing inspiration to improve novelty and

quality, the basic characteristics of creativity.

REFERENCES

Amabile, T. M. (1982). Social psychology of creativity: A

consensual assessment technique. Journal of Personality

and Social Psychology, 43(5), 997.

Chan, J., Siangliulue, P., Qori McDonald, D., Liu, R.,

Moradinezhad, R., Aman, S., Solovey, E. T., Gajos, K.

Z., & Dow, S. P. (2017). Semantically far inspirations

considered harmful? Accounting for cognitive states in

collaborative ideation. Proceedings of the 2017 ACM

SIGCHI Conference on Creativity and Cognition, 93–

105.

Colton, S., Goodwin, J., & Veale, T. (2012). Full-FACE

Poetry Generation. ICCC, 95–102.

Davis, N., Hsiao, C.-Pi., Singh, K. Y., Li, L., Moningi, S., &

Magerko, B. (2015). Drawing apprentice: An enactive

co-creative agent for artistic collaboration. Proceedings

of the 2015 ACM SIGCHI Conference on Creativity and

Cognition, 185–186.

Davis, N. M. (2013). Human-computer co-creativity:

Blending human and computational creativity. Ninth

Artificial Intelligence and Interactive Digital

Entertainment Conference.

Douglas, D. H., & Peucker, T. K. (1973). Algorithms for the

reduction of the number of points required to represent a

digitized line or its caricature. Cartographica: The

International Journal for Geographic Information and

Geovisualization, 10(2), 112–122.

Gatys, L. A., Ecker, A. S., & Bethge, M. (2015). A neural

algorithm of artistic style. ArXiv Preprint

ArXiv:1508.06576.

Gero, J. S. (1990). Design prototypes: A knowledge repre-

sentation schema for design. AI Magazine, 11(4), 26–26.

Gero, J. S., & Kannengiesser, U. (2004). The situated

function–behaviour–structure framework. Design

Studies, 25(4), 373–391.

Hoffman, G., & Weinberg, G. (2010). Gesture-based human-

robot jazz improvisation. 2010 IEEE International

Conference on Robotics and Automation, 582–587.

Jacob, M., Zook, A., & Magerko, B. (2013). Viewpoints AI:

Procedurally Representing and Reasoning about

Gestures. DiGRA Conference.

Jongejan, J., Rowley, H., Kawashima, T., Kim, J., & Fox-

Gieg, N. (2016). The quick, draw!-ai experiment. Mount

View, CA, Accessed Feb, 17, 2018.

Karimi, P., Grace, K., Maher, M. L., & Davis, N. (2018).

Evaluating creativity in computational co-creative

systems. ArXiv Preprint ArXiv:1807.09886.

Karimi, P., Maher, M. L., Davis, N., & Grace, K. (2019).

Deep Learning in a Computational Model for Conceptual

Shifts in a Co-Creative Design System. ArXiv Preprint

ArXiv:1906.10188.

Karimi, P., Rezwana, J., Siddiqui, S., Maher, M. L., &

Dehbozorgi, N. (2020). Creative sketching partner: An

analysis of human-AI co-creativity. Proceedings of the

25th International Conference on Intelligent User

Interfaces, 221–230.

Lucas, P., & Martinho, C. (2017). Stay Awhile and Listen to

3Buddy, a Co-creative Level Design Support Tool.

ICCC, 205–212.

Mamykina, L., Candy, L., & Edmonds, E. (2002).

Collaborative creativity. Communications of the ACM

45(10), 96–99.

Mikolov, T., Chen, K., Corrado, G., & Dean, J. (2013).

Efficient estimation of word representations in vector

space. ArXiv Preprint ArXiv:1301.3781.

Ramer, U. (1972). An iterative procedure for the polygonal

approximation of plane curves. Computer Graphics and

Image Processing, 1(3), 244–256.

Rehurek, R., & Sojka, P. (2010). Software framework for

topic modelling with large corpora. In Proceedings of the

LREC 2010 Workshop on New Challenges for NLP

Frameworks.

Reinig, B. A., Briggs, R. O., & Nunamaker, J. F. (2007). On

the measurement of ideation quality. Journal of

Management Information Systems, 23(4), 143–161.

Shah, J. J., Smith, S. M., & Vargas-Hernandez, N. (2003).

Metrics for measuring ideation effectiveness. Design

Studies, 24(2), 111–134.

Veale, T. (2014). Coming good and breaking bad: Generating

transformative character arcs for use in compelling

stories. Proceedings of the 5th International Conference

on Computational Creativity.

Yannakakis, G. N., Liapis, A., & Alexopoulos, C. (2014).

Mixed-initiative co-creativity.

CHIRA 2021 - 5th International Conference on Computer-Human Interaction Research and Applications

130