Is Anthropomorphic Design a Viable Way of Enhancing Interface Usability?

Chapter Three: The Web-Based Questionnaire

During the course of this research, various issues were raised concerning the application of ethopoeia (as described in Chapter One) and the way in which people might interact socially with their computers. In order to further examine these issues, a web-based questionnaire was constructed and administered. Information was gathered from different users and employed in an attempt to establish the extent to which they might anthropomorphise or personalise their own computers and to discover how they felt about existing humanlike interfaces, such as Clippy. Some questions were designed to ascertain which personality types users might prefer. This information was then used in the design of the anthropomorphic prototype implemented in the lab tests (described in Chapter Four).

3.1 Methods and Participants

The questionnaire was written in html, using Dreamweaver and also included a specially modified copy of a perl script which processed the results and submitted them via e-mail. Several methods were used to promote the questionnaire and everyone who took part was offered the chance to go into a draw to win a £20 book token as an incentive to complete it. The questionnaire was publicised at the University, via flyers which were handed out at the Project in Progress day (Poster Session). It was also advertised on fifty posters which were attached to noticeboards throughout the campus. In addition to this, a link to the questionnaire was distributed via personal e-mail, to friends and family, encouraging recipients to forward it on to ‘anyone who likes filling in questionnaires’. These methods of promotion proved very successful as 72 respondents completed the web form and returned their answers.

Of the 72 respondents, exactly half were male and half were female. 78% were under 35 years old, as most of the respondents were students or young professionals.

3.2 Results

3.2.1 How are People Using their Computers?

In order to better understand some of the answers submitted by respondents, the first seven questions were used to gather information on the type of user and established the context for their activities.

The majority of computer use amongst participants was stated as taking part at home, with work coming a close second. 54% of users described themselves as having intermediate computer skills, 35% rated their abilities as expert or almost expert.  80% of respondents used the computer for more than 11 hours a week – with academic or paid work stated as the main activity for 57% of respondents. (See fig.3.1)

3.2.2 How do People View their Computers?

As well as being questioned on any mindless or involuntary actions which they might apply, such as talking to their computer, users were also asked whether they have ever customised or personalised their machine. This information could give a further indication of the way in which a computer is perceived and offers an insight into the relationship between a user and their machine. For example, might someone who regarded their computer as a friend be more inclined to use special wallpaper of screensavers?

81% of users confirmed that they have personalised their computer with wallpaper. Of those, 90% described the wallpaper currently displayed on their desktop at that moment. The choice of pictures was very varied, ranging from a majority of personal images such as pictures of family and friends to more obscure items such as a tube map and a ‘weird monkey’. One important thing to note is the diversity of the choices. Of the 52 named items, 45 were all different.

In addition to wallpaper, users also customised their computers using various other methods (see Fig 3.2).

The next section of the questionnaire asked users how they regarded their computer and if they had ever interacted with it socially.

50% of users said that they had sometimes felt that the computer was against them, with 15% affirming that they often felt this way. 60% felt that the computer was their friend.

More than half of the users (53%) admitted that they talk kindly to their computer, with more women than men applying this approach. Conversely, more men than women acknowledged that they shout at their computer. 83% of users admitted to this practice.

The results showed that where social interaction took place between users and their computers, it was more likely to be verbal, than physical. 39% admitted that they had hit their computer at some point and 36% confessed to patting it fondly. (Fig. 3.3)

3.2.3 Talking Cars

As many texts on anthropomorphism cite computers and cars as examples of technology which are frequently regarded as humanlike, a question was also included asking users if they had ever spoken to their car, to see if there was a link between the two behaviours. 74% of users admitted to this. They were then questioned on whether they thought that KITT (from the Knight Rider TV series, mentioned in Chapter Two) was a good model for user interaction:

This question involved choosing the most applicable statement from a selection. 90% of respondents had seen Knight Rider, of these 38%  said that they would like to be able to converse with their car in this way, 25% said they thought that a talking car might be useful for navigation, but not just chat and 31% said that the car had kitsch entertainment value. 3% of respondents said that they would hate to be able to converse with their car and 3% said that they considered the notion of a talking car to be unrealistic

3.2.4 Reactions to the Office Assistant

95% of users said that they were familiar with Clippy (the Microsoft Office Assistant). When asked to describe Clippy in one word, 94% of those users offered an opinion. 37 different words or phrases where used to describe Clippy. Of these, only 5% were positive responses. 35% of people used the word ‘annoying’, 10% said Clippy was ‘irritating’.  Other negative comments ranged from ‘patronising’, ‘loathsome’, ‘rubbish’, ‘Satan’ and ‘smug’ to several expletives, which are unprintable here. Positive comments included ‘amusing’, ‘nice’ and ‘friendly’.

Users were then asked to choose an applicable statement regarding their overall perception of the Office Assistant, in order to ascertain if they shared similar views to the personal opinion on Clippy, expressed in Chapter Two (which describes how I liked it at first and then it began to irritate me).

Opinion seemed to be evenly spread between instant dislike and gradual disillusionment. 44% of users stated that they disliked Clippy immediately and have never used it.  43% professed that they had liked it at first, but then it started to annoy them. 4% said that they didn’t like it at first but then it ‘grew on them,’ and only 4% of users said that they found it really useful.

3.2.5 Choosing a Personality

As noted in Chapter Two, CASA theory (Nass, Steuer & Tauber, 1994) suggests that characters that are popular and likeable in the real world will also be more popular and likeable when used in a computer interface. Whereas previous research has investigated the use of personality types (such as dominant/submissive or introvert/extrovert) or personality traits (such as apologetic or humorous), for the purposes of this study it was decided to try a different approach, which has not been applied before. When considering the idea of someone who is ‘popular and likeable in the real world’ the idea of using celebrity personality traits seemed an incredibly obvious course of action, yet one which does not seem to have been previously explored.

Many people would like to ‘hang out’ with a celebrity - and pictures of television or film stars are a popular choice for computer wallpaper or screensavers. With this in mind, it was decided that the anthropomorphic prototype used in the lab tests should have a ‘celebrity personality.’ The final section of the questionnaire asked users if they would like to ‘hang out’ with a celebrity and asked their opinion on ten different celebrities which were being considered for use in the prototype.

The ten celebrities chosen were picked because they were considered to be both well known (in Britain) and potentially easy to characterise in an interface.

These results (Fig. 3.4) were then used to determine a ‘popularity score’ for each celebrity. This was calculated by adding together the number of people who ‘like’ or ‘don’t mind’ the celebrity and then subtracting the number of people who ‘dislike’ or were ‘unfamiliar’ with the celebrity.

For example:
18 people like David Beckham + 36 who don’t mind him = 54
14 people dislike him + 4 who are unfamiliar with him = 18
54 –18 = 36 …..so David Beckham’s popularity score is 36

The final question asked users which celebrity they would like to hang out with. It should be noted here that a third (33%) of the respondents said ‘no one’ – of the 48 remaining answers, 44 different celebrities were chosen, including sports stars (Johnny Wilkinson, Glenn Hoddle), pop stars (Robbie Williams, Dr Dre), actors (George Clooney, Julia Roberts) and television presenters (Jonathan Ross, Dermot O’Leary).

3.3 Discussion

3.3.1 Methodological concerns

Motivating such a large amount of people to fill in an undergraduate questionnaire was considered a relative success here. However, because of the way that the questionnaire was distributed and publicised, the participants have tended to be from a certain type of user group. Only two people aged 56 – 65 filled in the questionnaire, so it is not intended to be an accurate measure of a cross-section of society, but merely an indication, for the purposes of this small-scale study, of the type of issues which might concern users.

Chapter One of this study mentions how many test participants who are questioned after experiments emphatically deny the very behaviour that they have just exhibited (Reeves & Nass 1996). This may also apply to the people who responded to the questionnaire. Although many users freely  admitted that they had interacted socially with their computers, some people might not even realise they are doing so. For example I have only started consciously observing my own actions since I began this research – I have subsequently noticed that I only pat the computer when it is ‘in trouble’ and I only talk to the car when I have passengers (I consider that this may be because I am too busy shouting at other road users when I am driving alone).

3.3.2 The CASA Theory and Ethopoeia

This research certainly seems to support the findings of the CASA paradigm (Nass, Steuer & Tauber, 1994). With many users professing that they interact socially with their computer both verbally and physically and that they often regard it as a friend or enemy (or both).

Of the 53% of participants who admitted talking to their computer kindly, 90% also admitted that they speak to their car. Of the 47% who said that they did not talk to their computer kindly, only 68% claimed that they talk to their car. This may indicate that some users might be pre-disposed to talking to certain objects, whereas others are not.

Shouting at the computer was not included in the above analysis, as shouting may be caused by other factors apart from ethopoeia, such as basic frustration. In an exploratory study, users reported wasting an average of 42% - 43% of their time on a computer due to frustrating experiences (Lazar, Jones, Bessiere, Ceaparu & Shneiderman, 2003), so it is unsurprising that they might vent their frustration in this way. They might not necessarily be shouting at the computer as an interaction, but bellowing to relieve tension.

97% of users declared that they had personalised their computer in some way, so it would be difficult to tell if personalisation is linked to the relationship that they might have. Although it is worth noting that of the 60% who regarded their computer as a friend, 100% of those respondents also personalised their computer.

3.3.3 Reactions to the Talking Car

The questions relating to KITT showed that only 3% of users would hate to converse with their car. This illustrates that they might generally be receptive to the idea of an anthropomorphic, computerised car, especially if used in the right context. Though present technology is quite incapable of creating something as powerful as KITT, only 3% of users thought that the notion was unrealistic.

3.3.4 Reactions to Clippy

The respondents’ reactions to Clippy confirmed previously reported suggestions that most users dislike the Office Assistant. The diversity of disparaging remarks was breathtaking.

Users were also consistently diverse when describing their choice of other items, such as wallpaper or celebrity. Where invited to type in their own personal penchant, they expressed a broad range of preference, and rarely input the same item as each other. This may indicate that in order to please a greater number of people, more choice is required, combined with greater facilities for customisation.

3.3.5 The Celebrity Interface

One Questionnaire respondent e-mailed concerning the attribution of opinion to celebrities, stating that his own personal opinion did not involve ‘like or dislike’. He gave the example that he liked  Elvis’ music, but had little opinion of Elvis himself, similarly, with David Beckham, he said that he appreciated his footballing ability, but ‘not much beyond that’.

It was noted that although some people do have strong opinions of celebrities, others do not. As mentioned, 33% of respondents indicated that there was no particular celebrity that they would like to hang out with. Despite this, 67% of respondents indicated that they would  like to ‘hang out’ with a celebrity, so after some deliberation it was decided that this concept would still be used in the design of the anthropomorphic prototype, described in Chapter Four.

© Alison Flind 2006