TableOfContents

\section{Introduction}

The user's character is believed to influence the user's feeling of presence. The user's cultural background is often mentioned as such a characteristic \citep{IJsselsteijn+RidderETAL-Pres:00, Freeman+LessiterETAL-IntrPres:01}. A few cross-cultural presence studies are available \citep{Chang+WangETAL-Croscommmedilear:02}, but none investigated the relationship between the user's cultural background and presence directly. This influence is, at this point in time, more of a conjecture than a proven fact, and therefore we conducted an empirical study to investigate the relationship.

In absence of a clear definition of what cultural factors may influence presence, a good approach is to include participants from clearly different cultures. Using Dutch and Chinese participants in our study optimized cultural diversion. Hofstede \citep{Hofstede-Geert:91} provides an empirical framework of culture in which the Dutch and Chinese culture differ substantially.

At the same time, we were interested in distributed interactive media and their influences on the user's feeling of presence. We have entered a new media era: passive television programs become interactive with the red button on your remote control \citep{Bennett-ButtRevoPowePeop:04}. Video games come with many different controlling interfaces such as dancing mats, EyeToy\registered{} cameras, driving wheels and boxing Gametraks\texttrademark{} \citep{Games-Game:05}. The D-BOX\registered{} Odyssee\texttrademark{} motion simulation system even introduces realistic motion experiences, that were originally designer for theme parks, into our living rooms \citep{D-BOX-D-BOMotiSimu:05}. In the vision of Ambient Intelligence \citep{Aarts+Marzano-EverViewAmbiInte:03}, the next generation of people's interactive media experience will not unfold only on a computer or television, or in a head set, but in the whole physical environment. The environments involve multiple devices that enable natural interactions and adapt to the users and their needs.

Distribution is not a new idea for enhancing the entertainment experience. Multichannel surround sound systems distribute sound all around the audience and hence provide a more realistic and natural sound experience. The ambient intelligence \citep{Aarts+Marzano-EverViewAmbiInte:03} concept goes beyond such sound distributions by distributing content through other channels in the user's environment. Each display in the room may show video clips and each lamp may change its color and brightness. Especially intelligent appliances, such as domestic robots, may present synchronized behaviors that give the content a physical body.

However, the influence of embodiment on the user's presence experience seems unclear. On the one hand, embodiment extends the distributed content from an on-screen virtual environment to a physical environment. The physical embodiment improves the content's liveness and fidelity by stimulating more sensors of the user. This might result in an increased feeling of presence \citep{Lombart+Ditton-Hear:97}. On the other hand, the physical embodiment may transfer more attention from the virtual environment to the physical environment. This might break down the illusion of being there and hence would result in less feeling of presence \citep{Freeman+LessiterETAL-IntrPres:01}.

To control interactive content, the user requires interaction devices. A physical embodiment would invite direct manipulation. A robot could, for example, ask the use to touch its shoulder to select an option. Interaction with a virtual on-screen character may favor the use of a remote control. Embodiment in interactive media can therefore not be studied without considering the interaction method. We therefore included two interaction methods in our study.

In this framework of interactive distributed media we defined the following three research questions: \begin{enumerate}

\end{enumerate}

\section{Experiment}

We conducted a 2 (Interaction) $\times$ 2 (Embodiment) $\times$ 2 (Culture) mixed between/within experiment (see \fref{fig:TheInterview_conditions}). Interaction and culture were the between participant factors. Interaction had the conditions \expcondition{RemoteControl} and \expcondition{DirectTouch}, and culture had the conditions Dutch and Chinese. Embodiment was a within participant factor. Embodiment had the conditions \expcondition{ScreenAgent} and \expcondition{Robot}.

%attachment:figure1.png

\begin{figure}[hbtp]

  • \centering \includegraphics[width=10.5cm]{figures/TheInterview_conditions02} \caption{Conditions of The Interview experiment with Chinese and Dutch participants}%

    \label{fig:TheInterview_conditions}%

\end{figure}

\subsection{Measurements}

The questionnaire used was the ITC-SOPI \citep{Lessiter+FreemanETAL-CrosPresQues:01}. Only the definition of the Displayed Environment in the introduction was adjusted to include the robot/screen character. The questions remained unchanged and are clustered into four groups: \begin{inparaenum}

\end{inparaenum}

\subsection{Participants}

$19$ Chinese and $24$ Dutch between the age of $16$ and $48$ ($14$ female, $29$ male) participated in the experiment. Most of them were students and teachers from Eindhoven University of Technology, with various backgrounds in computer science, industrial design, electronic engineering, chemistry, mathematics and technology management. The Chinese participants were no longer than two years in the Netherlands.

\subsection{Setup}

The experiment took place in a living room laboratory (see \fref{fig:TheInterview_setup}). The participants were seated on a couch in front of a table. The coach was $3.5m$ away from the main screen, which was projected onto a wall in front. The projection had a size of $2.5m \times 1.88m$ with $1400 \times 1050$ pixels. The second screen was located $0.5m$ from the coach, standing on the table. The secondary screen was $30cm\times23cm$ with $1280 \times 1024$ pixels LCD touch screen (Philips DesXcape Smart Display).

%attachment:figure2.png

\begin{figure}[hbtp] \centering \subfigure[ScreenRemote/ScreenTouch]{ \includegraphics[width=5cm]{figures/TheInterview_setup_screen} } \subfigure[RobotRemote/RobotTouch]{ \includegraphics[width=5cm]{figures/TheInterview_setup_robot} } \caption{Experiment setup} \label{fig:TheInterview_setup} \end{figure} %

In the \expcondition{Robot} conditions, the secondary touch screen was replaced with a Lego robot that had about the same height. In the \expcondition{ScreenAgent} conditions, the secondary screen displayed a full screen agent of the robot.

The behavior of the screen based agent and the Lego robot were identical. They played the role of a TV companion by looking randomly at the user and the screen, but always looking at the user while speaking. Speakers were hidden under the table and were used to produce the speech, which was based on the standard Apple Speech Synthesis software. At the start of every movie, the character introduced himself and its role.

The interactive movie was about a job interview in which the participants had to make decisions for the applicant. The story and movie cuts were designed to be neither too exciting nor too boring. The movie had two decision points, which resulted in four possible movie endings. The participants chose different options almost all the time. At every decision point camera would zoom in on the applicant's forehead. The applicant then cycled through two options in his mind. He looked first to the left and thought aloud about one option, before he looked right and thought aloud about the second option. In the \expcondition{remote} condition the screen would show one icon on the left and a different icon on the right. The icons were identical to two icons on the remote control. In the \expcondition{robot} condition, the participant had to touch the left or the right shoulder of the robot to make the decision.

\subsection{Procedure}

After reading an introduction that explained the structure of the experiment the participants started with a training session. In this session, the participants watched an unrelated interactive movie that had only one decision point, during which the participants could make the decision using the remote control. Afterwards, they had the opportunity to ask questions about the process of the experiment. Next, the participant were randomly assigned to one of the between-participant conditions, which each consisted of two movies and a questionnaire after each movie. The participant received five Euros for their efforts.

\section{Results}

The mean scores for all measurements, including their standard deviations are presented in \fref{tab:TheInterview_mean_stddev} and graphically in \fref{fig:TheInterview_means}.

%attachment:table1.png

\renewcommand{\arraystretch}{1.2} \begin{table}[hbtp] \centering %\footnotesize \small \caption{Mean and standard deviation for all measurements} \label{tab:TheInterview_mean_stddev} \begin{tabular}{llllrr}\hline\hline Embodiment & Nationality & Interaction & Measurement & Mean & Std.Dev. \\ \hline \expcondition{ScreenAgent}& Chinese & \expcondition{RemoteControl} & Spatial Presence & 3.08 & 0.18 \\

\expcondition{Robot} & Chinese & \expcondition{RemoteControl} & Spatial Presence & 2.99 & 0.2 \\

\end{tabular} \end{table} \renewcommand{\arraystretch}{1}

%attachment:figure3.png

\begin{figure}[hbtp]

\end{figure}

\subsection{Embodiment, interaction and nationality effect}

A 2 (embodiment) $\times$ 2 (interaction) $\times$ 2 (culture) repeated measures ANOVA was conducted. Interaction had no significant influence on any of the measurements. Embodiment and culture both had significant influence on almost all measurements (see \fref{tab:TheInterview_embodiment}).

%attachment:table2.png

\begin{table}[hbtp] \centering %\footnotesize \small \caption{F and p values for culture and embodiment} \label{tab:TheInterview_embodiment} \begin{tabular}{llrr}\hline\hline Factor & Measurement & F (1,39) & p \\ \hline Embodiment & Spatial Presence & 4.789 & \bf 0.035 \\

Culture & Spatial Presence & 19.49 & \bf 0.001 \\

\end{tabular} \end{table}

Interaction was removed as a factor from the further analyses since it had no effect on the measurements. The means for all remaining conditions are summarized in \fref{fig:TheInterview_nationality_embodiment} and were used as the basis for the further analyses.

%attachment:figure4.png

\begin{figure}[hbtp]

\end{figure}

Paired Sample t-Tests were performed across both culture conditions. The measurements for Spatial Presence were significantly ($t(42)=2.235$, $p=0.031$) higher in the \expcondition{ScreenAgent} condition than in the \expcondition{Robot} condition. Negative Effects were significantly ($t(42)=2.38$, $p=0.022$) higher in the \expcondition{Robot} condition than in the \expcondition{ScreenAgent} condition.

Independent Samples t-Tests were performed. All measurements between the Dutch and the Chinese participants differed significantly, except for engage in the screen condition, which just missed the significance level ($t(41)=2.007$, $p=0.051$).

\section{Discussion}

The participants' cultural background clearly influenced the measurements. Chinese participants perceived more presence than Dutch participants in all conditions. One might suspect that the Chinese participants might simply be more polite in answering questions. Our measurements show that they also gave higher scores to Negative Effects and therefore did not simply respond politely. The next question will be what aspects of the cultural background have the greatest influence on presence. \Citet{Hofstede-Geert:91} suggested several categories through which cultures and organizations may be characterized, but none of them appear relevant to presence at first sight. One might speculate that the long-term orientation in Chinese culture would result in more patience towards imperfections. They might have more easily tolerated the noise emitted by the robot and the occasional visibility of a microphone in the movie. Further studies are necessary to investigate this issue.

The influence of embodiment on all measurements does not conform to the expected results defined in the construct of presence. If the user experiences high presence then all measurements should be high, including Negative Effects. Furthermore, Spatial Presence should be positively correlated with Negative Effects. However, in our results Spatial Presence and Naturalness are higher in the \expcondition{ScreenAgent} condition, while Negative Effects was higher in the \expcondition{Robot} condition. Negative Effects appear to have been affected by something else than presence.

During the experiment, the robot's motor emitted noise, which caused the participants to look at it. A moving physical object is potentially dangerous and hence attracts attention. Clearly, the robot emphasized the participants feeling of being in the room and not in the movie and thereby reducing the presence experience.

The participants were frequently switching between looking at the movie and the robot and hence divide their attention. This switching made it hard for the users to stay focused and might cause the high negative experience. \Citet{Eggen+FeijsETAL-BreaFlowIntecomp:03} showed that a divided attention space reduced the users immersion. Further research is necessary to determine if divided attention increases the negative effects of multiple displays. The extra costs necessary to build and maintain a robot for an interactive movie appear unjustified in relation to its benefit.

The different interaction methods (using a remote control or touching directly) had no influence on the measurements. The participants did not experience more or less presence when they interacted with a remote control or with the screen/robot directly. This is to some degree surprising, since the participants had to lean forward to touch the screen/robot directly, while they could remain leaned back using the remote. The necessity to make a choice might have overshadowed the difference in physical movement. To create compelling sense of presence it might be useful to pay more attention to the physical output than to the input.

\section{Acknowledgments}

The authors would like to thank Loe Feijs, Maddy Jansen and Kees Overbeeke for their help and support. In addition, we would like to thank all participants of the study.