Content uploaded by Thomas Tullis
Author content
All content in this area was uploaded by Thomas Tullis on Feb 16, 2017
Content may be subject to copyright.
Are People Drawn to Faces on
Webpages?
Abstract
Three studies were conducted to investigate the effects
of faces on webpages. In Study I, eye-tracking data
showed that users were clearly drawn to faces when
asked to look at pages and report what they remember.
In Study II, the presence of a face next to a message
on a webpage caused users to have a harder time
finding that message. In Study III, photos of the
authors of opinion articles caused users to be less likely
to find the article and to give the page worse ratings.
Keywords
Web, faces, pictures, photos, people
ACM Classification Keywords
H.5.4. Hypertext/Hypermedia
Introduction
Recognizing faces is important to people. Faces help us
distinguish friend from foe, family from non-family, and
Barack Obama from George Bush. How humans
recognize faces has been a research topic for
psychologists and biologists for many years. Some
studies have even shown that a specific part of the
brain (the Fusiform Face Area) is primarily devoted to
face recognition [1].
Copyright is held by the author/owner(s).
CHI 2009, April 4 – 9, 2009, Boston, Massachusetts, USA.
ACM 978-1-60558-247-4/09/04.
Tom Tullis
Fidelity Investments
82 Devonshire St., V3B
Boston, MA 02109 USA
tom.tullis@fmr.com
Marisa Siegel
Fidelity Investments
82 Devonshire St., V3B
Boston, MA 02109 USA
marisa.siegel@fmr.com
Emily Sun
Swarthmore College
500 College Avenue
Swarthmore, PA 19081 USA
esun1@swarthmore.edu
CHI 2009 ~ Spotlight on Works in Progress ~ Session 2
April 4-9, 2009 ~ Boston, MA, USA
4207
Given the importance of faces, it’s not surprising that
print advertisements have used pictures of people, and
especially their faces, for many years. This trend has
carried over to the web, where it’s hard to find an ad-
supported site that doesn’t include at least some ads
with pictures of people. Other common uses of faces on
the web include so-called “lifestyle photos” and photos
of “experts” accompanying their articles (Figure 1).
Figure 1. Samples of how faces are commonly used on the
web: in banner ads, “lifestyle” photos, and opinion pieces.
One aspect of pictures on the web that’s been
previously studied is whether photos of the people
associated with a website influence the perception of
trust of the site. The results have been somewhat
mixed. In one study of a prototype online banking site
[2], having photographs of employees of the company
(e.g., customer service representatives) increased
ratings of trust of the site. However, another study
using a modified version of the Amazon site [3], found
mixed effects of photos of the company’s staff. One
group of users responded positively to the photos while
three other groups did not.
Our interest in photos, and especially photos of obvious
faces, was quite simple: Are people somehow drawn to
the faces? If face recognition is almost “hard-wired”
into people, then do we have a strong tendency to be
attracted to photos of faces on the web? Their common
use in ads and other areas of the web would suggest
that at least some web designers think that faces do in
fact draw attention.
Study I
Our first study was an exploratory one designed to
learn more about what attracts visual attention on a
webpage. We used a Tobii eye-tracking system to
track where people look when initially viewing pages on
the Fidelity.com website. Twelve adults who use the
web on a daily basis participated in the study. Eight
had used the Fidelity.com site at some point in the past
while four had not. Each participant was shown static
images of 16 Fidelity.com pages for 7 seconds each.
They were instructed to “Scan the page as you would
when visiting a page for the first time.” They were also
told that after the page went away they would be asked
to briefly describe aloud the main purpose or topic of
the page and any key elements they remember. Of the
16 pages, 8 included faces and 8 did not.
For every page that included faces, they were among
the most-fixated elements of the page. Examples of
heatmaps for four of the pages containing faces are
shown in Figure 2. Even relatively small faces received
fixations.
CHI 2009 ~ Spotlight on Works in Progress ~ Session 2
April 4-9, 2009 ~ Boston, MA, USA
4208
Figure 2. Heatmaps showing the fixations on faces in
Experiment 1.
The obvious conclusion from this exploratory study was
that there might be something “special” about faces
that does in fact draw attention to them on webpages.
Study II
Study II built upon the findings of Study I through a
direct comparison of the effect of having a photo of a
face vs. no image. We expected a photo of a face to
draw attention, making tasks related to text adjacent to
it easier to complete than if there was no image.
Method
In this study, a section of a Portfolio Summary page
(Figure 3) was modified to contain either a photo of a
woman’s face or no image (Figure 4).
Figure 3. The Portfolio Summary page tested. The location of
the section that had two different treatments in Study II is
circled.
Figure 4. The section either included a photo of a woman’s
face or no image.
This was conducted as an online study. Participants
were randomly assigned to either the “Face” or “No
Image” condition and given the same six tasks to
complete in random order. Five of the tasks drew the
participant’s attention around the page, while one task
required the participant to look at the section in
question. All tasks had answers that could be found by
reading information on the page. None of the links were
CHI 2009 ~ Spotlight on Works in Progress ~ Session 2
April 4-9, 2009 ~ Boston, MA, USA
4209
active. Tasks were completed by selecting an answer
from a dropdown menu. Time to select the answer was
automatically recorded. At the conclusion of each task,
the participants were asked to rate task ease on a five-
point scale, and upon completing all tasks were asked
to rate the overall ease of finding information on a
seven-point scale.
Participants
413 Fidelity employees participated through an online
study received in the company’s daily email newsletter.
Additionally, 41 employees participated in the lab while
having their eyes tracked in the same manner
discussed in Study I.
Results
Data were analyzed for the task that required the
participants to find the critical section. The results are
shown in Figure 6. Participants in the Face condition
answered the critical task correctly significantly less
often than participants in the No Image condition (78%
correct vs. 93% correct, p<.01 by t-test). Additionally,
participants in the Face condition took significantly
longer to complete the task than those in the No
Image condition (54 sec vs. 37 sec, p<.01).
Participants in the Face condition rated the task
significantly harder than participants in the No Image
condition (3.4 vs. 4.2 on a 5-point scale of task ease,
p<.01). Finally, after doing all six tasks, participants in
the Face condition rated the overall ease of finding
information on the page significantly lower than
participants in the No Image condition (6.4 vs. 6.6 on a
7-point scale of Ease of Finding, p<.01). Analysis of
the fixation data collected in eye tracking indicated
that, in this context, participants rarely fixated on the
face (Figure 5). Contrary to our hypothesis, a picture
of a face in this context actually caused users to do
worse on a task involving information adjacent to the
face.
Accuracy (% Correct)
Face
No
Image
50%
60%
70%
80%
90%
100%
Mean Time (seconds)
No
Ima ge F ac e
20
25
30
35
40
45
50
55
60
Task Ease Rating
Face
No
Image
3.0
3.5
4.0
4.5
Ease of Finding
Information
Face
No
Ima g e
6.0
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
Figure 6. Data from Study II. Error bars represent the 90%
confidence interval of the means. Participants in the Face
condition were less likely to get the right answer, took longer,
rated the task as harder, and rated the page as harder to find
information on.
Study III
We hypothesized that the failure to find any benefit of
faces in Study II may have been due to the context in
which they were used. In Study III, we examined the
effect of an image of an “expert” accompanying the
introduction to his or her article. As illustrated
previously in Figure 1, this is a common practice on the
web.
Figure 5: Eye tracking data from
Study II showed that participants
rarely fixated on the photo of the
face. Although 78% of the
participants did eventually find the
answer, as indicated by the hot spot
on the text, it was in spite of the
presence of the face.
CHI 2009 ~ Spotlight on Works in Progress ~ Session 2
April 4-9, 2009 ~ Boston, MA, USA
4210
Method
A financial news page (Figure 7) was modified to
include photos of the authors of “Expert Opinion”
articles or to have no images. Two such articles were
on the page. In the Faces condition, actual photos of
the authors (one male and one female) were shown
(Figure 8). In the No Images condition, the photos
were not included.
Figure 7. The financial news page used in Study III. The two
“Expert Insight” articles that either included a photo or not
were on the right (and shown larger in Figure 8).
Similar to Study II, this was conducted as an online
study. The same six tasks were used for both
conditions, of which two required the user to find the
key articles. All tasks required the user to click on a
link which displayed an “answer number”. Different
links yielded different answer numbers. The accuracy
of task completion was determined using these answer
numbers. Time to select the answer was automatically
recorded. Upon completing each task, participants were
asked to rate their confidence in their answer.
Additionally, upon completing all tasks, participants
were asked to rate their trust in the accuracy of the
information on the web page.
Participants
333 Fidelity employees participated in the online study
which was posted in the company’s daily email
newsletter. They were randomly assigned to either the
Faces or No Images condition.
Results
As in Study II, data analysis was done for the two
critical tasks. Results are shown in Figure 9.
Accuracy (% Correct)
No
Ima g e Faces
75%
80%
85%
90%
95%
100%
Mean Time (seconds)
No
Image Faces
25
30
35
40
45
Confidence in Answer
No
Image Faces
4.0
4.2
4.4
4.6
4.8
Trust
No
Image Faces
5.60
5.80
6.00
6.20
6.40
Figure 9. Data from Study III. Error bars represent the 90%
confidence interval of the means. Participants in the Faces
condition were less likely to get the right answer, took
marginally longer, were less confident in their answers, and
rated their trust in the accuracy of the information lower.
Figure 8. The two
versions of the “Expert
Insight” articles, with and
without photos of the
authors.
CHI 2009 ~ Spotlight on Works in Progress ~ Session 2
April 4-9, 2009 ~ Boston, MA, USA
4211
Participants in the Faces condition were significantly
less likely to get the correct answers than those in the
No Image condition (87% vs. 94% correct, p=.03).
Participants in the Faces condition were also less
confident in their answers than the participants in the
No Image condition (4.4 vs. 4.7 on a 5-point scale of
confidence, p=.02). Finally, participants in the Faces
condition trusted the accuracy of the information
less than participants in the No Image condition (6.0
vs. 6.3 on a 7-point scale of trust in the accuracy,
p=.05). Although the difference in the amount of time
to complete the tasks did not reach significance, there
was a tendency for participants in the Faces condition
to take longer than those in the No Image condition (40
sec vs. 36 sec, p=.19). These results indicate that
including the faces not only caused users to do worse
on the tasks, it also caused them to feel less confident
in their answers and trust the accuracy of the
information less.
Discussion and Conclusions
The results of Studies II and III were quite surprising to
us. Not only did the presence of faces not help in
either study, they actually had a negative impact on
performance related to information associated with the
faces, and they had a negative impact on subjective
reactions. We believe this may largely be due to the
over-exposure many web users have had to faces in
ads on the web. Some users may have simply learned
to “tune out” faces. However, we also found that faces
had less of a negative impact on task performance in
Study III compared to Study II (a 7% accuracy
difference in Study III vs. a 15% accuracy difference in
Study II). This indicates that the context of the faces
clearly makes a difference, and the “Expert Opinion”
context of Study III was more appropriate.
The following conclusions appear to be supported by
these studies:
As shown in Study I, there are situations where
people are drawn to faces on webpages. However,
these situations may be the exception rather than the
rule. The artificial conditions of that study may only
generalize to a case where someone is browsing a
website with no particular task in mind.
The more task-oriented users are, the less likely
they are to believe that the information they’re looking
for would be associated with a photo of a face, at least
for the contexts of Studies II and III.
Future Directions
Future research in this area should investigate different
user types, including factors such as age, gender,
geographic location, and web experience. Specific
comparisons between browsing and task-based
activities may help explain the disparity between the
results of Study I and Studies II and III.
References
[1] Cell Press (2006, April 9). Explaining How The
Brain Recognizes Faces. ScienceDaily. Retrieved
January 6, 2009, from http://www.sciencedaily.com-
/releases/2006/04/060409153926.htm.
[2] Steinbrueck, U., Schaumburg, H., Duda, S., and
Krueger, T., A, Picture Says More Than A Thousand
Words - Photographs As Trust Builders In E-Commerce
Websites. Proceedings of CHI2002: Extended Abstracts.
Minneapolis, MN, US, April 20-25, 2002, pp. 748-749.
[3] J. Riegelsberger & M. A. Sasse (2002). Face it -
Photos don’t make a Web Site Trustworthy. Extended
Abstracts CHI2002, 20-25 April, Minneapolis, MN, pp.
742-743.
CHI 2009 ~ Spotlight on Works in Progress ~ Session 2
April 4-9, 2009 ~ Boston, MA, USA
4212
