EyeChess: the tutoring game with visual attentive interface Špakov Oleg Department of Computer Sciences University of Tampere Finland March, 2005 AAFG 2005

O.Špakov1/ The number of gaze-aware interfaces increases rapidly. Part of them, namely gaze-controlled interfaces, allow to employ the gaze as an input. The majority of interaction procedures involves interface object selection. The best known gaze-based selection methods are following: Dwell time Gaze-gesture Blink However, novice users of such interfaces may be unaware of how to use their gaze to reach the best performance. Introduction EyeChess

Some researchers used primitive games to train participants to control the gaze before they were involved in testing the gaze- controlled interface [Gips, 1996] [Nelson, 1992]. Games could be used also as a tool to evaluate efficiency of an interaction methods There are no real gaze-based games on the market available yet. Due to the problems of eye-tracking devices, the gaze-aware interfaces must deal with the objects large enough to eliminate the accuracy gap. Cell-based games seems ideal for this purpose. O.Špakov2/ Introduction EyeChess

O.Špakov3/ Goal EyeChess The goal of this project was to develop a PC-based tutorial to assist novices in playing chess endgames and to carry out a pilot evaluation of the efficiency of the proposed technique. Chess game was chosen for this project as the one, which satisfies all conditions mentioned above. Moreover, this game looks promising from other research aspects (for example, for studying the processes of mental activity). The game supposes to be used as a tutorial of endgames by the players with no or very little experiences in playing chess.

3.Motion validity (red/green 3D) 4.Motion target (light-yellow) O.Špakov4/ Software design - Feedback EyeChess Visual feedbacks (static): 1.What the user is looking at (3D) 2.Motion target (dark-yellow)

Visual feedbacks (dynamic): Animation of the piece movement Sound feedback Alert sound at the attempt to move to forbidden square Congratulations as reward when the task was completed O.Špakov5/ Software design - Feedback EyeChess

O.Špakov7/ Software design - Tutoring EyeChess The tutoring is supporting by flashing 3D border of the square, which is supposed to be selected next, after the player made 2 attempts to make a wrong move. The tutorial does not support wrong movements The flashing border has either the piece to move or the destination square. It depends on whether the player found the correct piece to move or not If the player has found the correct piece to move due to flashing 3D border, the destination square will flash only when s/he makes an attempt of wrong move once more.

The game is playing in full-screen mode. It was implemented as a plug-in for iComponent application [O. Špakov’s home page] O.Špakov8/ Software design - Tutoring EyeChess The flashing appears only when the correct square is in focus (i.e. is observed by the player) The algorithm to support the flashing of the square to be selected is presented on the figure

O.Špakov9/ Pilot study EyeChess The aim of the pilot study was to test, what gaze-based selection method could be the most suitable for this game. The following parameters and selection methods were tested: Dwell time = 1.8 s. Gesture: gaze moves from the object to select to any off-screen button (placed in every corner of the monitor), and back to the same object. Blink with duration from 350 to 1000 ms. Players and procedure: four subjects were asked to move any 10 pieces they would like to any position, using each selection method. Selection time was measured and the players’ opinions were collected during the pilot test

O.Špakov10/ Pilot study EyeChess Hardware: iView X RED III eye tracker from SensoMotoric Instruments Results (the average time and player rating): Dwell time: 3.3 s, quite OK for this task Gesture: 2.8 s, less convenient, especially for permanent selection Blink: 2.5 s, worst, very annoying The first selection method was chosen. However, the second one (gesture) is still could be used since in a real chess game the selection itself does not occur very often

The subjects: three females and one male, who did not participate in the pilot test and have no or very little experience in playing chess. Two of them had some experience using eye trackers, and 2 others had not. Hardware: Tobii 1750 eye tracker from Tobii Technologies Procedure: after the subjects read the instruction how to use tutorial, they were asked to play “whites” and solve 20 endgames in 3 moves. Before the test started, they were trained to use the environment solving 2 other endgames. Completion time and the number of wrong attempts were measured during the test O.Špakov11/ Method EyeChess

O.Špakov12/ Results EyeChess The average time was 71.4 s. (σ=19.3). By moves: 1.56 s. (78.4%) s (13.4%) s. (8.1%) Errors rate: 18% of all attempts (σ=6%). By moves: 34% 6% 1%

O.Špakov13/ Results EyeChess Incorrect piece in 8% of all attempts (σ=4%). By moves: 1.21%2. 1%3. 1% Subjects often selected a wrong piece (23% overall), but then changed their mind and did not move it. By moves: 1.58%2. 5%3. 6% It seems that after the first move was performed, the subjects did mistakes or select wrong pieces only because of the calibration drift.

O.Špakov14/ Results EyeChess In almost every tenth task (9%) the subjects attempted to make a wrong move at least 2 times and all of them found the hint (blinking square). Nobody made 2 or more wrong attempts after the first correct move was found The subjects were stressed during first 7 tasks and they analyzed the given task carefully. This resulted to a long thinking. Then they relaxed and the number of error increased (trials 9-13). At the end all subjects solved tasks faster than at the beginning. The wrong attempts after the first successful move were occurred rarely.

O.Špakov15/ Conclusion EyeChess EyeChess is a PC-based gaze-controlled tutorial to assist novices in playing endgames. To make a move, the player first selects a piece and then its destination square. The visual and sound effects provide a good level of interaction. The tutorial provides a hint (flashing highlight) in case then the player is unaware of the correct move. The pilot study revealed that dwell time was the preferred gaze- based selection method.

O.Špakov16/ Conclusion EyeChess At the end of the test no significant improvements in the gaze interaction were found: both groups of subjects showed the similar trial completion time at the beginning and end of the test. The following study revealed that the tutorial was helpful in guiding the decision-making process. However, the tasks were too easy and the EyeChess tutoring support was rarely used. Overall, the game is eye tracker-friendly and thus easy to play, as the pieces and squares on the chessboard are large enough to accommodate calibration drifts.

O.Špakov17/ Future work EyeChess The project will be continued to support both novices and experts. The author plans also to extend the game to a full- range game with an opponent: a computer or another player over the net.

O.Špakov18/ Gips, J., Dimattia, P., Curran, F.X. and Olivieri, P., Using EagleEyes – an electrodes based device for controlling the computer with your eyes – to help people with special needs, in J. Klaus, E. Auff, W. Kremser & W. Zagler (eds.), Interdisciplinary Aspects on Computers Helping People with Special Needs, Oldenburg, pp , 1996 Nielsen, J., Noncommand User Interfaces. Communications of ACM 36, pp , 1992 O. Špakov’s home page: iComponent application, at References EyeChess