Understanding Worldwide Human Information Needs

Revealing Cultural Influences in HCI by Analyzing Big Data in Interactions

Rüdiger Heimgärtner

HCI Research & Development, Intercultural User Interface Consulting (IUIC), Lindenstraße 9, 93152 Undorf, Germany

Keywords: Culture, Relation, Structure Equation Model, Neural Networks, Factor Analysis, Data Analysis, Global,

Local, Intercultural Interaction Analysis, Tools, Big Data, Modeling, Data Mining, Statistics.

Abstract: Understanding human information needs worldwide requires the analysis of much data and adequate

statistical analysis methods. Factor Analysis and Structural Equation Models (SEM) are a means to reveal

structures in data. Data from empirical studies found in literature regarding cultural human computer

interaction (HCI) was analyzed using these methods to develop a model of culturally influenced HCI. There

are significant differences in HCI style depending on the cultural imprint of the user. Having knowledge

about the relationship between culture and HCI using this model, the local human information needs can be

predicted for a worldwide scope.

1 INTRODUCTION

Rapidly progressing globalization requires

consequent adaptation of the methods and processes

applied in Human Computer Interaction (HCI)

design to the relevant cultural needs. Intercultural

User Interface Design (IUID) integrates several

disciplines. For example it integrates information

technology with cultural studies within this

prominent and complex field. However, the relations

between HCI and culture are not yet well elaborated,

even if there are some initial approaches (cf. e.g.,

Röse et al., 2001); (Marcus, 2006); (Vatrapu and

Suthers, 2007); (Clemmensen, 2009); (Heimgärtner,

2012). A proper taxonomy of all approaches,

methods and processes in IUID is also still missing,

even if there are some clues (cf. e.g., Clemmensen

and Röse, 2010). In addition, there are several tools

for user interaction logging (e.g. ObSys for

recording and visualization of windows messages,

cf. Gellner and Forbrig, 2003). However, none of the

existing tools provide explicitly cultural usability

metrics to measure culturally imprinted interaction

behavior. This is because the connections between

HCI and culture are not systematically collected and

prepared for such tools.

This paper describes some methods to elucidate

the connection between HCI and culture by

analyzing big data (cf. Auinger et al., 2011): the

intercultural interaction analysis tool (IIA tool),

neural networks, factor analysis and structural

equation models (SEM). Then, the application of the

methods to analyze the connections between culture

and HCI are explained and the results and challenges

are addressed.

2 METHODS FOR ANALYZING

CULTURE IN HCI

The following methods and tools can be used to

yield cultural differences in HCI and to determine

the relationship between culture and HCI. Cultures

are orientation systems for group members (cf.

Thomas et al., 2010). The characteristics of cultures

can be described using cultural dimensions (cf.

Hofstede and Hofstede, 2005). The interaction

behavior of the user with the system can be

described with HCI dimensions (cf. Heimgärtner

2012). The aim is to determine the connection

between cultural dimensions and HCI dimensions

and its values using adequate methods and tools to

yield a model for culturally influenced HCI that

serves to predict the HCI style of members of any

cultures. With this knowledge, relevant design

recommendation can be derived to develop user

interfaces with high usability. Quantitative values

for the indices of the cultural dimensions are

available from empirical studies by Hofstede and

Hofstede 2005. Further quantitative values for the

165

Heimgärtner R..

Understanding Worldwide Human Information Needs - Revealing Cultural Inﬂuences in HCI by Analyzing Big Data in Interactions.

DOI: 10.5220/0004166401650169

In Proceedings of the International Conference on Data Technologies and Applications (DATA-2012), pages 165-169

ISBN: 978-989-8565-18-1

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

indicators of the HCI dimensions are also available

(cf. Heimgärtner, 2012). 916 valid data sets together

with the values of the indices of Hofstede are used to

determine the connections between the values of the

cultural and HCI dimensions.

2.1 IIA Tool as a Method Framework

A special tool for measuring the interaction behavior

of the user has been developed by the author (cf.

Heimgärtner, 2008). The interaction of culturally

different users doing the same test task can be

observed (using the same test conditions i.e. the

same hard- and software, environment conditions,

language and the test tasks) as well as requiring the

same experience with the use of the system. Logging

data of dialogs, debugging and HCI event triggering

while using the system are highly valuable

(Kralisch, 2006). This data can be logged during

usability tests according to certain user tasks. The

IIA tool provides data collection, data analysis and

data evaluation. It serves to record and analyze the

user’s interaction with the system in order to identify

culturally influenced variables such as color,

positioning, information density and interaction

speed as well as their values. The collection and

preparation of the data is to be done automatically

for the most part by the IIA data collection module.

The data is to be stored in databases in a format that

is immediately usable by the IIA data analysis

module, which does subsequent data conversion or

preparation. Common statistic programs like SPSS

or AMOS can be deployed to apply statistical

methods (cf. Ho, 2006). The IIA data evaluation

module enables classification using neural networks

to cross-validate the results from data analysis.

The Delphi-IDE allows transformation of new

HCI concepts very quickly into well formed

prototypes that can be tested very soon within the

development process. For example, some hypotheses

have been confirmed quantitatively addressing many

test users online in one month (implementing the use

cases as well as doing data collection and data

analysis). Hence, using the IIA tool means rapid use

case design, i.e. real-time prototyping of user

interfaces for different cultures as well as a very

large amount of valid data collected quickly and

easily worldwide online via internet or intranet or

offline locally on the spot.

2.2 Using Neural Networks

Neural networks are used within the IIA evaluation

module to verify and establish trends of cultural

differences in user interaction and to enhance the

plausibility of quantitative results. For example, it

might not be important which subjects take part in a

test, if neural networks are used, which can

independently learn existing trends (e.g., back

propagation networks (cf. Haykin, 2008) or self-

organizing maps (cf. Kohonen, 2001)). Therefore,

such networks are not concerned about which test

persons take part in the test. By connecting the

categorized grouped test data according to the HCI

dimensions to the input neurons and the cultural

characteristics (represented for example by variables

like nationality, mother tongue, etc.) of the users to

the output neurons of the neural network, training of

the network will reveal if there is a correlation of the

values of the HCI dimensions with the input and the

culture with the output of the neural network. In

other words, if cultural differences do exist, i.e. if

there is a correlation between the corresponding test

data of the test persons and the culture at the output

of the neural network, the neural network will learn

and reveal it (“monitored learning”, cf. Mandl et al.,

2003). By means of connecting test data which is

categorized or grouped according to hypotheses and

the cultural variables to the output of the neural

network, it identifies whether cultural differences do

exist and to which degree. Thereby, it can be seen

whether correlations exist between the test data from

the subject with the input of the neural network and

his cultural imprint with the output of the neural

network.

2.3 Explorative Factor Analysis

HCI dimensions represent classes of indicators. For

example, the class “number of information units per

space unit” belongs to the HCI dimension

“information density” and can be expressed by the

indicator “number of words displayed on a screen”.

Another HCI dimension is “interaction frequency”.

This dimension contains the class “number of

interactions per time unit” (e.g., represented by the

indicator “number of mouse clicks per second”).

Support for the correctness of the HCI dimensions

comes from the application of factor analysis

methods.

Explorative factor analysis serves hereby to

derive the HCI dimensions by grouping potential

variables to factors (component matrix) according to

the strength of their correlation (correlation matrix)

(cf. Ho, 2006). Using main component analysis with

the 916 data sets including 19 indicators, four main

components have been extracted (ranked by impact,

cf. Heimgärtner, 2012): (i) Style of user interaction

DATA2012-InternationalConferenceonDataTechnologiesandApplications

166

behavior (expressed by interaction speed index and

interaction exactness index) is influenced by age,

time abroad and experience with the use of

computers as well as task orientation of the user. (ii)

Information reception (represented by information

density, number and order index) varies because of

the influence of uncertainty avoidance. (iii)

Information speed (represented by the information

speed index) correlates with help usage (represented

by the number of help calls). Furthermore, (iv)

gender and power distance indices seem to be

correlated. The four main factor loadings

(“components”) derived from the data explain nearly

95% of the variance. The remaining 15 components

in total explain only 5% of the variance. This

analysis is clear, if not very meaningful because the

extensions of the indicators constituting the

components partly overlap. Therefore, an

“extended” (more detailed) factor analysis has been

carried out. It revealed that component #1

(interaction behavior) in the “simple” factor analysis

can be split up into the two components “primary

cultural imprint” and “interactional behavior” or

“HCI style”. Hence, the largest components concern

HCI and cultural aspects, which now can be further

used to analyze their connections doing structural

equation modeling.

2.4 Structural Equation Modeling

Structural equation models (SEM) belong to the

statistical methods of confirmative factor analysis,

which can be performed using e.g., AMOS (cf.

Byrne, 2001). A structural equation model consists

of a set of equations. The effect or endogenous

variable is on the left side and on the right side is the

sum of the causes with each causal variable

multiplied by a causal parameter. Here, structural

equation modeling serves to identify the

relationships between cultural dimensions and HCI

dimensions and their causes. SEM has, on the left

hand side the cultural indices by Hofstede

representing cultural dimensions and on the right

hand side, the indicators of the HCI dimensions.

Cultural dimensions and HCI dimensions are

connected by assumed relations in the SEM. The

theory is the better the more variances in the

empirical data can be explained statistically by the

SEM. The theory depending on the modeled

relations is best if the balance between the variables

is in equilibrium. Adding or removing variables or

relations to improve the equilibrium is called

structural equation modeling. First attempts using

AMOS showed that the higher the relationship

orientation (collectivism) represented by Hofstede’s

individualism index (IDV), the higher the

information density, information speed, information

frequency, interaction frequency and interaction

speed as well as the other way around. Additional

methods like conformational factor analysis as well

as regression analysis support this process of finding

the right model.

3 REVEALING CULTURAL

INFLUENCES IN HCI

3.1 Results

The data analyzed for the qualitative and

quantitative studies revealed a trend for the

investigated cultures that allowed shifting towards a

model of culturally influenced HCI. (cf.

Heimgärtner, 2012). With the right combination of

indicators representing the HCI dimensions, it is

possible to capture interaction differences that are

culturally imprinted (e.g., according to cultural

aspects such as nationality, mother tongue, country

of birth, etc. or to cultural dimensions). There are

correlations between the interaction of the users with

the system and their cultural background. The

cultural differences in HCI concern layout (complex

vs. simple), information density (from high to low),

personalization (from greater to lesser), language

(symbols vs. characters), interaction speed (from

high to low) and interaction frequency (from high to

low). In addition, the cultural differences found in

HCI by discriminance analysis are quantitatively

measurable by a computer system using a special

combination of indicators represented by interaction

patterns depending on the culturally imprinted

interaction behavior of the user. The recognition and

classification of cultural inter-action patterns in HCI,

i.e. cultural differences in HCI, can be achieved

purely quantitatively (cf. Heimgärtner, 2012). A

handful of indicators are sufficient for this purpose.

Moreover, the interaction patterns representing the

cultural differences in HCI and the derived

indicators are statistically sufficiently discriminating

to enable computer systems to detect them

automatically and to assign the users to a certain

cultural imprint. Furthermore, when reversed,

interaction patterns are also useful to identify a user,

which is necessary for the system to adapt to the

corresponding user.

UnderstandingWorldwideHumanInformationNeeds-RevealingCulturalInfluencesinHCIbyAnalyzingBigDatain

Interactions

167

3.2 Challenges

However, many open issues still remain. The

reliability of the test equipment and test methods as

well as the results and models must be discussed and

alternative test settings in the future should be used.

The results have been determined statistically and

are mainly descriptive. Explaining the inference

statistics for the evidence and the reasons of the

cultural differences in depth remains to be done. In

addition, there is the need for strengthening the

confirmation of the HCI dimensions by conducting

deeper explorative factor analysis with the data from

further studies as well as enhancing the separation

effect and discriminatory power of the indicators and

their classes. The test data sets must be evaluated in

more detail to generate optimized algorithms for

cultural adaptability in HCI based on neural

networks, which need large amounts of interaction

data for training, validating and testing as well as on

structured equal models to prove basic theoretical

and well explained interaction models by taking

cultural aspects into account.

4 CONCLUSIONS

Many kinds of culturally influenced interaction

patterns are only recognizable over time requiring

the collection of big data. Hence, enhanced

algorithms and tools must be used for data analysis.

Therefore, designing tools for non-experts to do

their own analysis with big data in interactions will

be a prominent task for interaction designers in the

future (cf. Fisher et al., 2012). The combination of

different statistical methods to determine cultural

differences and influences in HCI represents an

initial idea and the first step to ascertain the right

relationships between culture and HCI. However,

much effort still remains. Nevertheless, the

presented approach and model are worthy of being

investigated and optimized in the future. Revealing

cultural influences in HCI by analyzing big data in

interactions to finally create and use a model or even

a theory for culturally influenced HCI should help to

better understand human information needs

worldwide.

REFERENCES

Auinger, A., Aistleithner, A., Kindermann, H., Holzinger,

A., 2011. Conformity with User Expectations on the

Web: Are There Cultural Differences for Design

Principles? In: Marcus, A. (Ed.) Design, User

Experience, and Usability. Theory, Methods, Tools

and Practice. Lecture Notes in Computer Science

LNCS 6769/2011. Springer. Heidelberg. Berlin. New

York. 3-12.

Byrne, B. M., 2001. Structural equation modeling with

AMOS: Basic concepts, applications, and

programming, Lawrence Erlbaum. Mahwah, NJ.

Clemmensen, T., 2009. Towards a Theory of Cultural

Usability: A Comparison of ADA and CMU Theory.

In: HCD 2009, Held as Part of HCI International

2009, San Diego, CA, USA, July 19-24.

Clemmensen, T., Röse, K., 2010. An Overview of a

Decade of Journal Publications about Culture and

Human-Computer Interaction (HCI). In: Human Work

Interaction Design: Usability in Social, Cultural and

Organization Contexts. IFIP Advances in Information

and Communication Technology, Volume 316/2010,

98-112.

Fisher, D., DeLine, R., Czerwinski, M., Drucker, S., 2012.

Interactions with big data analytics. In: Interactions,

Volume 19, 3 (May 2012). ACM. 50-59.

Gellner, M., Forbrig, P., 2003. ObSys–a Tool for

Visualizing Usability Evaluation Patterns with

Mousemaps. In: Jacko, J., Stephanidis, C., (Eds.),

Human-Computer Interaction, Vol. I, Theory and

Practice, Proceedings of the 10th International

Conference on HCI. Lawrence Erlbaum Associates.

Mahwah, New Jersey, London. 469-473.

Haykin, S., 2008. Neural networks: a comprehensive

foundation, Prentice Hall.

Heimgärtner, R., 2008. A Tool for Getting Cultural

Differences in HCI. In: K. Asai, Human Computer

Interaction: New Developments. InTech. Rijeka. 343-

368.

Heimgärtner, R., 2012. Cultural Differences in Human

Computer Interaction - Towards Culturally Adaptive

Human Machine Interaction. Dissertation. University

of Regensburg. Oldenbourg. München.

Ho, R., 2006. Handbook of univariate and multivariate

data analysis and interpretation with SPSS, Chapman

& Hall/CRC. Boca Raton, Fla.

Hofstede, G., Hofstede, J. G., 2005. Cultures and

Organizations: Software of the Mind. McGraw-Hill.

New York, USA, 2

Edition.

Kohonen, T., 2001. Self-organizing maps, Springer.

Berlin.

Kralisch, A., 2006. The Impact of Culture and Language

on the Use of the Internet Empirical Analyses of

Behaviour and Attitudes, Dissertation. Berlin.

Mandl, T., Schudnagis, M., et al., 2003. A Framework for

Dynamic Adaptation in Information Systems. In: C.

Stephandis. Human Computer Interaction: Theory and

Practice. Proceedings of 10th International

Conference on Human-Computer Interaction, 22-27

June 2003. Crete. 425-429.

Marcus, A., 2006. Cross-Cultural User-Experience

Design. In: Diagrammatic Representation and

Inference.16-24.

Röse, K., Zühlke, D., et al. 2001. Similarities and

DATA2012-InternationalConferenceonDataTechnologiesandApplications

168

Dissimilarities of German and Chinese Users. In: G.

Johannsen. Preprints of 8th IFAC/IFIP/IFORS/IEA

Symposium on Analysis, Design, and Evaluation of

Human-Machine Systems. Germany. Kassel. 24-29.

Thomas, A., Kinast, E.-U., Schroll-Machl, S., 2010 (Eds.).

Handbook of intercultural communication and

cooperation. Basics and areas of application.

Vandenhoeck & Ruprecht. Göttingen.

Vatrapu, R., Suthers, D., 2007. Culture and Computers: A

review of the concept of culture and implications for

intercultural collaborative online learning. In: Lecture

Notes in Computer Science 4568. Springer. 260-275.

UnderstandingWorldwideHumanInformationNeeds-RevealingCulturalInfluencesinHCIbyAnalyzingBigDatain

Interactions

169