MAR Shopping Assistant Usage: Delay, Error, and Utility

Subhashini Ganapathy, Glen J. Anderson, Igor V. Kozintsev Intel Corporation

EXTENDED ABSTRACT

This poster will present findings from a study of a shopping assistant prototype with simulated augmented reality information. The goal of the study was to find out the acceptable level of delay in presentation of augmented information and the acceptable rate of error of the information presented. Twelve participants interacted with a Samsung Omnia™ smartphone that presented a wine shopping scenario under several levels of delay in showing product information. Participants indicated their willingness to wait for each delay they experienced. Participants also answered a survey about which types of products they would want a shopping assistant application to assist with.

Participants were recruited with the following characteristics: • Household income of over $25K • Currently use their phone for at least two activities other

than voice (e.g., texting, accessing the web, taking photos, etc.)

• Navigated an unfamiliar city within the last 12 months • Experienced with street navigation applications such as

Google Maps or Mapquest • Majority had used a street navigation device like Neverlost

or Garmin • Interested in having access to a street navigation feature on

their phone or other mobile device (4 or 5 on 5-pt scale) • Interested in having access to a shopping feature on their

phone or mobile device that allows them to quickly find reviews and price comparisons on an item they’re considering (4 or 5 on 5-pt scale)

• Ages 25 to 49, half of each gender Participants had just finished participating in another study on

MAR information presentation for a tourist assistant prototype, so they already had some experience with a simulated MAR environment.

A simulated scenario of wine shopping was presented to participants on a Samsung Omnia™ phone (3.2 in./81mm diagonal screen, WQVGA, 400 x 240 pixels at 145 ppi). By taking a picture of a wine bottle, the application on the phone would match the image of the bottle with a database listing and present related information, such as competitive pricing and reviews.

Test sessions proceeded as follows: 1. In one-on-one sessions, the facilitator explained how the

shopping assistant on such a device would work and what kind of information it would present.

2. The participant then held the phone and saw the image in Figure 1 and was told this would be the view approaching the wine display in a hypothetical store.

3. The facilitator instructed the participant to tap the screen to advance to the next image, which appears in Figure 2.

4. Participants were then told to press the “Take Picture” button to simulate capturing an image in order to retrieve additional information about the bottle of wine in the middle of the image.

5. The image in Figure 3 then appeared, with the wheel in the image spinning to show the search was in progress.

6. After a variable period of time, the image in Figure 4 would appear.

7. Participants then indicated whether they thought the time they had to wait was acceptable.

Figure 1. Image of wine display

Figure 2. Bottle selected to capture in an image

Figure 3. Animation during wait period

Figure 4. Wine information appears

207

IEEE Virtual Reality 201119 - 23 March, Singapore978-1-4577-0038-5/11/$26.00 ©2011 IEEE

Each participant experienced up to 7 levels of delay before wine information appeared: .5 second, 1 second, 3 seconds, 5 seconds, 8 seconds, 12 seconds, and 18 seconds. In order to reduce the influence of learning effects, half the participants experienced the delay in ascending order and half in descending order. In the case of ascending order, the facilitator stopped showing higher delay levels once the participant said the delay had reached an unacceptable level.

Most participants indicated that acceptable delay occurred between 5 and 8 seconds. Participants who experienced ascending wait times showed less patience for delay, tending to cluster closer to 5 seconds, while participants experiencing descending delay indicated they would allow for a time closer to 8 seconds (see Figure 5).

Figure 5. Acceptable maximum delay

In follow-up discussion, participants indicated that the

acceptable delay depends on the importance of the item being shopped for. They said they would wait longer for items that were expensive. Some said they would not use it as often in a spontaneous way if the delay was too long, just when they were making a planned purchase. Some participants questioned the validity of results after a very short delay of 1 second or less.

Example quotes:

At 18 seconds: “This seems like it’s taking a really long time.”

At 12 seconds: “Okay, that was two foot taps…. If I know how long it takes, I can adjust my expectations.” “That was way too long. I wouldn’t use it as much.”

At 3 seconds: “Ooh, that was good. That was like the perfect speed.”

At .5 seconds: “That’s almost too fast.” “I don’t know how it could ever find it that fast. It must’ve

searched that wine before.” After indicating acceptable delay, the facilitator showed the

participant an errant condition in which the wrong information was retrieved for the bottle of wine. Most participants indicated

they would tolerate up to a 5% incidence in errors and still find the application useful.

Finally the participants answered a survey on importance of a shopping assistant supporting various types of products. The rating scale was as follows:

5-essential to have 4-would be very good 3-would be nice but not important 2-does not matter 1-would not want it Figure 6 shows the results of the importance survey.

Confidence intervals for the means appear as white bars. Some of the middle-importance items had higher variance and thus wider confidence intervals. For example, participants were more split on the importance of cookware, meats, and fresh produce.

Mean Importance

Figure 6. Results of importance survey

Getting information on expensive items was most important to participants. Other important items were those that exposed a purchaser’s taste or judgment, such as wine. Personal products and medicines were rated highly since they could affect health. Most grocery items were of low interest to most participants, but some were interested in getting information on aspects of food such as whether it was organic and where it was grown or raised. A few also mentioned that information about cheeses could be helpful, since the taste of artisan cheeses can be difficult for the inexperienced to predict.

KEYWORDS: Mobile augmented reality, shopping, latency,

delay, error, usage model.

INDEX TERMS: H.1.2 [Information Systems, MODELS AND PRINCIPLES, User/Machine Systems, Human factors]

REFERENCES

[1] J. Allebach. Binary display of images when spot size exceeds step size. Applied Optics, 15:2513–2519, August 1980.

[2] E. Catmull. A tutorial on compensation tables. In Computer Graphics, volume 13, pages 1–7. ACM SIGGRAPH, 1979.

[3] Peter Litwinowicz and Lance Williams. Animating images with drawings. In Andrew Glassner, editor, Proceedings of SIGGRAPH ’94 (Orlando, Florida, July 24–29, 1994),Com- puter Graphics Proceedings, Annual Conference Series, pages 409–412. ACM SIGGRAPH, ACM Press, July 1994.

208