EFFECTIVENESS AND PREFERENCES OF

ANTHROPOMORPHIC USER INTERFACE FEEDBACK IN A PC

BUILDING CONTEXT AND COGNITIVE LOAD

Pietro Murano, Christopher Ede

University of Salford, Computer Science, Multimedia and Telecommunications

Centre for Virtual Environments, Newton Building, Gt. Manchester, M5 4WT, U.K.

Patrik O’Brian Holt

Interactive Systems Research Group, School of Computing, The Robert Gordon University

St. Andrew Street, Aberdeen, AB25 1HG, Scotland

Keywords: Anthropomorphism, User interface feedback, Evaluation, Cognitive Load.

Abstract: This paper describes an experiment and its results concerning research that has been going on for a number

of years in the area of anthropomorphic user interface feedback. The main aims of the research have been to

examine the effectiveness and user satisfaction of anthropomorphic feedback in various domains. The

results are of use to all interactive systems designers, particularly when dealing with issues of user interface

feedback design. Currently the work in the area of anthropomorphic feedback does not have any global

conclusions concerning its effectiveness and user satisfaction capabilities. This research is investigating

finding a way for reaching some global conclusions concerning this type of feedback. The experiment

detailed, concerns the specific software domain of software for in-depth learning in the specific context of

PC building. Anthropomorphic feedback was compared against an equivalent non-anthropomorphic

feedback. The results were not statistically significant to suggest one type of feedback was better than the

other. It was also the aim to examine the types of feedback in relation to Cognitive Load Theory. The results

suggest that the feedback types did not negatively affect Cognitive Load.

1 INTRODUCTION

The user interface is usually the most visible part of

a software application and is often one of the main

aspects that users really think about if they have a

problem or if they like something about the

application they are using. It is therefore very

important to strive to achieve as good quality a user

interface as possible as this will affect the users’

perceptions about a system. In the worst case, if the

user interface is unusable, the whole application

could be abandoned.

The global aim of this research is to constantly

discover better ways of developing user interfaces,

specifically the feedback that is given to users. The

research is directly concentrating on investigating

the effectiveness and user satisfaction of

anthropomorphic feedback. To achieve this direct

comparisons are being made with non-

anthropomorphic feedback in an experimental

setting. Furthermore, the authors of this paper are

also trying to explain the results of conducted

experiments in terms of appropriate theories. One

such theory that is being investigated in conjunction

with the experimental results is the Cognitive Load

Theory.

Anthropomorphism at the user interface usually

involves some part of the user interface, taking on

some human quality (De Angeli, Johnson, and

Coventry, 2001). Some examples include a synthetic

character acting as an assistant or a video clip of a

human.

1.1 Background Literature

Other researchers have conducted some

investigations into the area of anthropomorphic

181

Murano P., Ede C. and O’Brian Holt P. (2008).

EFFECTIVENESS AND PREFERENCES OF ANTHROPOMORPHIC USER INTERFACE FEEDBACK IN A PC BUILDING CONTEXT AND COGNITIVE

LOAD.

In Proceedings of the Tenth International Conference on Enterprise Information Systems - HCI, pages 181-188

DOI: 10.5220/0001687301810188

 SciTePress

feedback, but the wide spread results of these efforts

do not reveal an overall global picture indicating if

such types of feedback are preferable (or not). One

of the earliest studies into anthropomorphism at the

user interface was by Quintanar, Crowell, Pryor and

Adamopoulos (1982). This was an experiment in a

quiz context which tested anthropomorphic textual

feedback and non-anthropomorphic textual

feedback. The quiz was about ‘psychology’ and the

participant sample was undergraduate students. The

main aims of the study were to obtain the user’s

thoughts about the system and ascertain the

effectiveness of the two types of feedback. One

finding of the study showed that participants

perceived the anthropomorphic feedback to be ‘more

human, less honest and slightly less courteous…’

compared with the non-anthropomorphic feedback.

Also Quintanar et al (1982) found the

anthropomorphic textual feedback to be more

effective. This was based on the fact that the quiz

scores were higher under the anthropomorphic

condition and the amount of time spent thinking

about the questions presented and the system’s

responses was higher. While this was an interesting

study, it did not address other forms of

anthropomorphism, such as synthetic characters or

video clips etc. Also some methodological aspects

could have caused some bias in the results. One such

issue concerns the recruitment of the participants.

The authors state that participants were screened for

their psychology knowledge, which is appropriate.

However they do not state how the different

experimental conditions were balanced in terms of

the participants’ psychology skills.

Also, Moreno, Mayer and Lester (2000) have

done some work in relation to anthropomorphic user

interfaces. Two experiments described in (Moreno et

al, 2000) looked at varying the kind of

communication an agent used towards the user. With

such variations they wanted to know if participants

in a certain experimental group could have ‘deeper

understanding’. They also wanted to discover if deep

understanding would be affected if using voice

compared to text. The first study varied the agents in

four different conditions in a botany context. The

first condition used a synthetic type agent able to

‘converse’. The second condition used only a

‘conversing’ agent (i.e. no image or animation). The

third condition used a synthetic type agent

communicating by means of text only. The fourth

condition used only text. The second study used the

same four conditions. However the synthetic agent

and its corresponding synthetic ‘voice’ were

replaced with a real human. Their results show that

the presence of an on screen image of an agent did

not significantly affect learning. However they

obtained significant results in favour of using a

voice to communicate information compared with

text. Their results suggest that using a voice helps to

improve learning. They found experimental

participants’ ability to remember and problem

solving skills to be improved. Also the voice agent

was rated more highly by experimental participants.

However some experimental design flaws suggest

that more work is required in this area. Their

publication (Moreno et al, 2000) does not detail

clearly enough if the participants were screened

properly for their prior botany knowledge. Some

screening did take place, but how this was done is

not detailed and could have biased the results if

some participants, e.g. happened to have more

botany knowledge in one condition. Also the paper

does not detail the level of difficulty of the botany

material used in the context of the experiment in

relation to the participants’ experience.

A further study concerning tutoring by

Moundridou and Virvou (2002) tested 2 conditions

in an algebra tutoring environment. The participants

were screened in advance for their mathematical

knowledge by means of a test and were deemed to

be approximately equivalent to one another. The

first experimental condition had a talking synthetic

face and the second was the same as the first

condition with text replacing the synthetic face. The

main results showed that there was no significant

difference between the 2 conditions for task time

completion. However the participants in the

anthropomorphic condition enjoyed the experience

more, found the system more useful and less

difficult to use. Lastly the participants were given a

post-experiment test and this did not reveal any

statistically significant difference between the

conditions in terms of overall results.

Further, the authors of this paper have been

investigating the appropriate use of anthropomorphic

feedback for some time and the results obtained are

not definitive with respect to obtaining global

knowledge in this area, e.g. (see Murano, 2005,

2003, 2002a, 2002b, 2001a and 2001b). However

related to the experiment reported in this paper, the

study by Murano (2002b) in the same domain of

online learning and the specific context of English

pronunciation, showed with significant results that

using an anthropomorphic feedback was more

effective and preferred by users. This in effect meant

that the anthropomorphic condition aided the

correction process more. This was an experiment

which used Italian participants with imperfect

ICEIS 2008 - International Conference on Enterprise Information Systems

182

English pronunciation. Several tasks involving

pronunciation exercises were used where either an

anthropomorphic or non-anthropomorphic feedback

was used to assist in the correction process. The

anthropomorphic feedback consisted of a video of a

human and the non-anthropomorphic feedback

consisted of guiding text and a diagram.

2 PC BUILDING EXPERIMENT

2.1 Aims and Objectives

Therefore this paper investigates further the domain

of online learning. This time it is in the specific

context of PC building. The aim of this experiment

was to gather data regarding effectiveness and user

satisfaction in the PC building context. Specifically

the aim was to find out if anthropomorphic user

interface feedback fostered a better interaction

experience with fewer errors and therefore a better

task completion rate. It was also of interest to find

out if anthropomorphic user interface feedback led

to better user satisfaction.

The authors were also interested to find out if

cognitive load was a factor in some of the results

observed. This theory essentially argues that when

the overall cognitive load exceeds a particular

threshold, then activities such as learning are

impaired or become more difficult (Martin-

Michiellot and Mendelsohn, 2000). Furthermore

cognitive load has three basic strands. The first is

‘intrinsic cognitive load’. This concerns the activity

involved in learning some item of information and

typically how many units of information are being

learned at the same time. The second strand

concerns ‘extraneous cognitive load’. This has to do

with how learning materials are presented to a

human. Therefore the more complex the manner of

presentation is, the higher the extraneous cognitive

load tends to be (Martin-Michiellot and Mendelsohn,

2000). The third strand concerns ‘germane cognitive

load’. This strand has to do with the human faculty

of processing and understanding information, and

problem solving (Sweller, van Merrienboer and

Paas, 1998).

2.2 Users

• All the participants taking part in the study were

of varied age groups.

• 30 participants were used in the experiment.

Although gender was not the main aspect under

consideration the sample used was

approximately 50:50 for gender.

• The participants had varied occupations.

• All the participants were novices to the area of

PC assembly. This ascertained by administering

a small pre-experiment test and only those with

low scores, (i.e. little knowledge about hardware

etc) were used.

2.3 Experimental Design

A between users design was used. The 30

participants were randomly assigned to one of the 2

conditions being tested –anthropomorphic or non-

anthropomorphic. Randomness was achieved by

alternately assigning a participant to one of the

conditions until all 30 had been assigned.

2.4 Variables

The independent variables were the types of

feedback, i.e.:

• Textual instructions.

• MS Agent synthetic character.

The dependent variables were the participants’

performance in carrying out the tasks and their

subjective opinions.

The dependent measures were that the

performance was measured by counting the number

of errors made as each participant attempted to

assemble the components, whether the participant

completed a task and the time taken to complete a

task. These factors were then used in a scoring

formula in order to achieve a single score per

participant (see note below). The formula was

devised so as to allow groups of similar types of

errors to be catalogued under a dedicated category.

The errors were determined by the experimenter

physically examining the way the components had

been inserted into the PC case (see note below). The

time taken to complete the tasks and the number of

times a participant clicked a certain button in the

application were recoded automatically by the

application. The subjective opinions were measured

by means of a post-experiment questionnaire

(Appendix 23).

(NOTE – The formula used was as follows:

• Each participant (unknown to them) was

started on 5 points for each task.

• For a completed task with 1 minor error, 1

point was deducted. A minor error was of

the kind that led to a device working, but

still having some problem, e.g. not securing

the CD ROM drive to the case with screws.

EFFECTIVENESS AND PREFERENCES OF ANTHROPOMORPHIC USER INTERFACE FEEDBACK IN A PC

BUILDING CONTEXT AND COGNITIVE LOAD

183

• For a completed task with 1 major error, 2

points were deducted. A major error was of

the kind leading to a device not working

but essentially still being fitted in place,

e.g. fitting the CD ROM drive

appropriately but not being able to insert

the required power cable.

• For a completed task with 1 severe error, 3

points were deducted. A severe error was 2

or more major and minor errors e.g. not

securing the CD ROM drive to the case

with screws was a minor error and not

being able to insert the required power

cable was a major error.

• For an unsuccessful attempt at completing a

task, 4 points were deducted, e.g. a

participant tried to do a task, but then gave

up or a participant tried to do a task, but

made 2 or more severe errors.

• If no attempt at all was made for a task, all

5 points were deducted to leave a score of

0.)

2.5 Apparatus and Materials

• A PC running Windows XP with 256 Mb RAM.

• Microsoft Agent 2.0 ActiveX component.

Lernout and Hauspie TruVoice Text-To-Speech

engine. The prototype was engineered with

VB.NET.

• External speakers.

• Desktop microphone used for the

anthropomorphic condition.

• An open PC case with the motherboard,

processor and power supply already assembled.

• Disassembled RAM board, CD ROM drive,

ribbon cable and relevant assembly screws.

• Screw driver.

2.6 Procedure and Tasks

The first step was to recruit a suitable number of

participants particularly meeting the requirement of

being novices to the assembly of computer

hardware. As stated above this was achieved by the

participants completing a pre-experiment

questionnaire and a small pre-experiment test

covering basic knowledge of PC components (only

participants with a low score in the pre-experiment

test were used – a high score would have indicated

too much knowledge for the experiment). Each

participant was briefed with the following points

before commencing the actual experiment:

• The software was developed for evaluating its

suitability to teaching about PC building.

• Help could be received from the system by using

the help button (non-anthropomorphic group) or

by asking for help via the microphone

(anthropomorphic group).

• The information for each stage would be shown

for a limited period of time. If the information

was to be repeated, this could be achieved by

pressing the ‘repeat’ button (non-

anthropomorphic group) or by asking for the

information to be repeated via the microphone

(anthropomorphic group).

• Video demonstrations were available for viewing

concerning the assembly of each part.

• It would not be possible to backtrack to a

previous stage on screen.

• A post-experiment questionnaire would need to

be completed at the end of the experiment.

Then the procedure described below was carried

out in the same way for all participants using the

same environment, equipment and

questionnaires/observation protocols. Each

participant was treated in the same manner. This was

all in an effort to control any confounding variables.

There were 2 tasks involving PC building.

Specifically, the first task concerned inserting a

RAM board into its appropriate slot in the

motherboard. The second task involved assembling a

CD ROM drive into the case, with a correct master

jumper setting and connecting the necessary cables

and screws.

Each participant was booked an appointment

during the day. The experiment took about 30

minutes to complete per volunteer. After completing

the initial pre-experiment questionnaire and test,

participants were able to view the PC case and

motherboard information screens. This allowed the

participant to become familiar with the software

before taking part in the assembly tasks.

Then for the first task, the participant was

directed to press the ‘Random Access Memory

Walkthrough button’. This would then initiate a

series of steps, with accompanying photographs of

the parts etc, describing what had to be done to

complete the task. For the anthropomorphic

condition, the Merlin character would narrate the

various aspects required and where relevant would

move on the screen and ‘point’ to certain elements

on the photographs. For the non-anthropomorphic

condition the same content was delivered textually

and to match the Merlin character’s movement and

pointing to various elements on the photographs,

clearly visible arrows were used to ‘point’ to the

ICEIS 2008 - International Conference on Enterprise Information Systems

184

same elements on the screen. The textual

information was displayed on the screen for the

same amount of time as the Merlin character took to

narrate the information. However the participants in

each condition had the option of having the

information repeated whilst within a particular

information stage.

Once the information had been received the

participant would attempt to physically insert the

RAM into the appropriate slot. Although the actual

information regarding the RAM board could not be

viewed again at this stage, the participant did have

the software’s help available. Also the information

regarding the PC case and motherboard which had

been made available before starting the first task of

inserting the RAM was available at this stage. Lastly

the relevant video demonstration was playable by

the participant at this stage. Once the first task was

completed, the second task was undertaken in the

same manner as described above.

At all times during the experiment the

participants were informally observed and at the end

of a task, an inspection of the relevant components

and their positioning etc. was carried out by the

experimenter – any errors being noted.

Lastly the participants were asked to complete a

post-experiment questionnaire regarding their

subjective opinions of their experience with the

software.

2.7 Results

The data collected was analysed using MANOVA

analysis. Regarding effectiveness, the amount of

time taken for each task was recorded, the number of

errors committed were recorded and also if a task

was completed was recorded. For user satisfaction,

various subjective opinions were elicited from the

participants.

As discussed in section 2.4 above, the factors of

task completion and errors were used in a scoring

formula. The results for the first task involving the

RAM board are presented below in Table 2.1. The

analysis involved the actual scores obtained, the

experimental group, the age groups of the

participants and the gender of the participants.

The F-ratio of 1.14 shows that there are no

significant differences between the various factors

analysed.

Also the same analysis was carried out for the

second task involving the fitting of a CD ROM

drive. This is shown in Table 2.2 below. The

analysis involved the actual scores obtained, the

experimental group, the age groups of the

participants and the gender of the participants.

Table 2.1: Analysis of Variance, Task 1, Overall Score.

Source DF Sum of

Squares

Mean Square F Ratio

Model 5 5.161467 1.03229 1.1362

Error 24 21.805200 0.90855 Prob > F

C. Total 29 26.966667 0.3685

Table 2.2: Analysis of Variance, Task 2, Overall Score.

Source DF Sum of

Squares

Mean Square F Ratio

Model 5 10.994271 2.19885 1.8008

Error 24 29.305729 1.22107 Prob > F

C. Total 29 40.300000 0.1508

Table 2.3: Analysis of Variance, Task 1, Times.

Source DF Sum of

Squares

Mean Square F Ratio

Model 5 67156.03 13431.2 2.4924

Error 24 129330.63 5388.8 Prob > F

C. Total 29 196486.67 0.0593

Table 2.4: LSMeans Differences Student's t.

Alpha = 0.050 t = 2.0639LSMean[i] By LSMean[j].

Mean[i]-Mean[j]

Std Err Dif

Lower CL Dif

Upper CL Dif

female male

female 0

82.7179

26.9818

27.0302

138.406

male -82.718

26.9818

-138.41

-27.03

Level Least Sq Mean

female A 387.47313

male B 304.75522

Levels not connected by same letter are significantly different.

Table 2.5: Analysis of Variance, Task 2, Times.

Source DF Sum of

Squares

Mean Square F Ratio

Model 5 54857.62 10971.5 0.6896

Error 24 381815.34 15909.0 Prob > F

C. Total 29 436672.97 0.6361

As with the previous task, the F-ratio of 1.80

shows that there are no significant differences

between the various factors analysed.

EFFECTIVENESS AND PREFERENCES OF ANTHROPOMORPHIC USER INTERFACE FEEDBACK IN A PC

BUILDING CONTEXT AND COGNITIVE LOAD

185

The tasks were also timed and included in the

analysis. For the first task (RAM Board insertion)

the analysis involved the actual times obtained, the

experimental group, the age groups of the

participants and the gender of the participants. This

is shown in table 2.3 above.

The F-ratio of 2.49* is tending towards

significance (p < 0.05). The initial results were then

subjected to post-hoc testing using a t-test, where

significance was shown for the gender, where

female participants were significantly slower than

the male participants. This can be seen in Table 2.4

above.

For the second task (CD ROM fitting) the

analysis also involved the actual times obtained, the

experimental group, the age groups of the

participants and the gender of the participants. This

is shown in table 2.5 above.

The F-ratio for the task involving the times

shows no significance for the various factors being

analysed.

Regarding the participant subjective responses

various aspects were considered. The main aspect of

interest concerns the application being considered

helpful for PC assembly. This was in relation to the

Table 2.6: Analysis of Variance, Application Considered

Helpful.

Source DF Sum of

Squares

Mean Square F Ratio

Model 5 6.979297 1.39586 2.1175

Error 24 15.820703 0.65920 Prob > F

C. Total 29 22.800000 0.0981

Table 2.7: LSMeans Differences Student's t.

Alpha = 0.050, t = 2.0639LSMean[i] By LSMean[j].

Mean[i]-Mean[j]

Std Err Dif

Lower CL Dif

Upper CL Dif

Anthropomorphic Non-

Anthropomorphic

Anthropomorphic 0

0.81598

0.32733

0.1404

1.49157

Non-

Anthropomorphic

-0.816

0.32733

-1.4916

-0.1404

Level Least Sq Mean

Anthropmorphic A 7.7819451

Non-Anthropomorphic B 6.9659605

Levels not connected by same letter are significantly different.

Table 2.8: Analysis of Variance, Application Frustrating

to Use.

Source DF Sum of

Squares

Mean Square F Ratio

Model 5 18.272845 3.65457 3.3699

Error 24 26.027155 1.08446 Prob > F

C. Total 29 44.300000 0.0191

experimental group, age group and gender. This is

shown in table 2.6 above.

Strictly the F-ratio of 2.12 is only approaching

significance (p < 0.05). However the initial results

were then subjected to post-hoc testing using a t-test,

where significance was shown for the experimental

group, where the anthropomorphic group

significantly considered the application more helpful

than their counterparts in the non-anthropomorphic

group. This can be seen in Table 2.7 above.

A further aspect of interest concerns the

application being frustrating to use. This was in

relation to the experimental group, age group and

gender. This is shown in table 2.8 above.

This result is significant (p < 0.05) with an F-

ratio of 3.37*, particularly in relation to the

experimental groups. The non-anthropomorphic

group significantly rated the application as more

frustrating to use compared to their counterparts in

the other anthropomorphic condition.

2.8 Conclusions

As can be seen in the previous section various

aspects were statistically analysed. Regarding the

effectiveness issues, no statistical significance can

be seen in the scores of the two groups which

included the task completion successes and the

errors made by participants.

The times were also analysed and for the first

task, there was no statistical significance for the

experimental group. However there was statistical

significance to show that the female participants

were slower than the male participants. Although

this research is not primarily about gender, it would

have been interesting to know why this result is in

this direction. Unfortunately none of the female (or

any of the participants) participants were available

again for interview. Also the initial demographic

data that was collected as part of the recruitment

process does not reveal any information that could

enlighten the authors on the matter. However the

authors believe that it is possible that something in

the female group’s background could shed light on

the matter. Further, for the second task, there is no

statistical significance for the time taken to complete

ICEIS 2008 - International Conference on Enterprise Information Systems

186

the task. This is for the experimental group and the

gender.

The user preference issues analysed suggest that

overall the preferences tended towards the

anthropomorphic condition. Participants in the

anthropomorphic condition rated the application as

more helpful than the participants in the non-

anthropomorphic condition. Also the non-

anthropomorphic group significantly rated the

application as more frustrating to use than the

anthropomorphic condition participants. Despite the

statistical analysis indicating a user preference

towards the anthropomorphic feedback, the authors

suggest caution in categorically declaring the

anthropomorphic feedback as being more satisfying

to use. This is because as described in section 2.6

above, the non-anthropomorphic condition had the

feature incorporated where the textual information

remained on the screen only for a certain amount of

time. This aspect was designed into the system to

more closely match the fact that the

anthropomorphic feedback condition, consisting of a

character with accompanying speech bubbles, only

appeared on the screen for the time it took the

character to utter the information. In essence the

authors did not want the non-anthropomorphic

condition participants having an unfair advantage by

having the information available to them for a much

longer period of time. While it is argued that this

feature should have balanced the two conditions

more closely, the authors suggest that this

potentially incurred the side effect of participants in

the non-anthropomorphic group rating the

helpfulness of the application significantly lower

than the other participants in the other experimental

group and rating the application as significantly

more frustrating to use.

Comparing the results for this experiment with

the results of previous work carried out by Murano

(2002b), it is clear that the effectiveness issues do

not match because in the English pronunciation

context the anthropomorphic feedback was more

effective, while in the experiment detailed above

there were no significant differences in the two

tested conditions. The only slight agreement

between the two experiments concerns the

subjective opinions of the participants. In both cases

it appears that participants tend to prefer the

anthropomorphic feedback given the similar domain

and contexts being considered. However as argued

in the previous paragraph, the subjective responses

for the experiment described above need to be

tempered with the knowledge that the attempts at

having well balanced experimental conditions may

have caused more negative opinions being given

towards the non-anthropomorphic feedback.

3 THE EXPERIMENT AND

COGNITIVE LOAD

As stated in the introduction of the paper, it was also

of interest to know if cognitive load could have been

a factor somewhere in the conditions and user

interface being tested. Informally, the fact that there

were no significant differences in effectiveness

could indicate that the cognitive load was equivalent

under both conditions. However the following

discussion in relation to cognitive load should help

to make things clear.

Firstly the participants were definitely beginners

to PC assembly as indicated by the recruitment

process. This characteristic could mean there could

have been some intrinsic cognitive load as they were

all learning something new to them with a few units

of information. However this would have spanned

both conditions and not been affected by the

feedbacks being tested, so therefore should have

been the same under both conditions. Regarding the

possibility of extraneous cognitive load, the very

nature of the character moving on the screen could

have increased this form of cognitive load, as

perhaps more integration on the part of the user

would have been needed, compared with the static

textual explanations which were definitely in

integrated format. The possibility of one condition

incurring a higher germane cognitive load is

unlikely, as the content of the information was the

same in both conditions. Also one could argue that

the textual information was static and perhaps gave

more advantage to the user compared to the Merlin

character’s speech and speech bubbles. However this

was not the case as the textual information was not

on the screen all of the time as discussed above.

Therefore if cognitive load issues had been at play, it

would have been expected that participants in the

non-anthropomorphic condition should have overall

had significantly more success in the task – this was

not the case though. Lastly this experiment timed the

participants whilst they did the tasks and generally

no significant result was observed between the

conditions (except for female participants). The

reason this is also of interest is that cognitive load

has been linked to task completion times as an

indicator of increased cognitive load (Neerincx, van

den Dobbelsteen, Grootjen and van Veenendaal,

EFFECTIVENESS AND PREFERENCES OF ANTHROPOMORPHIC USER INTERFACE FEEDBACK IN A PC

BUILDING CONTEXT AND COGNITIVE LOAD

187

2003 and Wang, Kaufman, Mendoca, Seol, Johnson

and Cimino, 2002).

It can therefore be argued that the two conditions

being tested were approximately equivalent in terms

of cognitive load. As stated, if cognitive load had

been a significant factor in one of the feedbacks, it

would have been expected that the performance of

participants in the anthropomorphic condition should

have had worse scores and worse overall times for

task completion.

3.1 Overall Conclusions

As can be seen by the results of this experiment and

also the previous work briefly considered in this

paper, there is still work to be done in this area to try

and determine the suitability of such types of

feedback. More work is being carried out. Results

are being further statistically analysed and also

analysed in terms of other theories of cognition and

human processing.

ACKNOWLEDGEMENTS

The authors would like to thank for their support:

The School of Computing Science and Engineering,

University of Salford. Heriot-Watt University,

Edinburgh, Dept. of Computer Science. The Robert

Gordon University, Aberdeen, School of

Computing.

REFERENCES

De Angeli, A, Johnson, G. I. and Coventry, L., 2001. The

Unfriendly User: Exploring Social Reactions to

Chatterbots, Proceedings of the International

Conference on Affective Human Factors Design,

Asean Academic Press.

Martin-Michiellot, S. and Mendelsohn, P., 2000.

Cognitive Load While Learning With a Graphical

Computer Interface, Journal of Computer Assisted

Learning, 16, 284-293, Blackwell Science.

Moreno, R., Mayer, R. E. and Lester, J. C. , 2000. Life-

Like Pedagogical Agents in Constructivist Multimedia

Environments: Cognitive Consequences of their

Interaction. ED-MEDIA 2000 Proceedings, p. 741-

746. AACE Press.

Moundridou, M. and Virvou, M., 2002 Evaluating the

Persona Effect of an Interface Agent in a Tutoring

System. Journal of Computer Assisted Learning, 18, p.

253-261. Blackwell Science.

Murano, P., 2005. Why Anthropomorphic User Interface

Feedback Can be Effective and Preferred by Users, 7th

International Conference on enterprise Information

Systems, Miami, USA, 25-28 May 2005. (c) -

INSTICC.

Murano, P., 2003. Anthropomorphic Vs Non-

Anthropomorphic Software Interface Feedback for

Online Factual Delivery 7th International Conference

on Information Visualisation (IV 2003) An

International Conference on Computer Visualisation

and Graphics Applications, London, England, 16-18

July 2003, (c) - IEEE.

Murano, P., 2002a. Anthropomorphic Vs Non-

Anthropomorphic Software Interface Feedback for

Online Systems Usage, 7th European Research

Consortium for Informatics and Mathematics

(ERCIM) Workshop - 'User Interfaces for All' -

Special Theme: 'Universal

Access'.Paris(Chantilly),France 24,25 October 2002.

Published in Lecture Notes in Computer Science (C) -

Springer.

Murano, P., 2002b. Effectiveness of Mapping Human-

Oriented Information to Feedback From a Software

Interface, Proceedings of the 24th International

Conference on Information Technology Interfaces,

Cavtat, Croatia, 24-27 June 2002.

Murano, P., 2001a. A New Software Agent 'Learning'

Algorithm, People in Control An International

Conference on Human Interfaces in Control Rooms,

Cockpits and Command Centres, UMIST, UK, IEE.

Murano, P., 2001b. Mapping Human-Oriented

Information to Software Agents for Online Systems

Usage, People in Control An International Conference

on Human Interfaces in Control Rooms, Cockpits and

Command Centres, UMIST, UK, IEE.

Neerincx, M. A., van den Dobbelsteen, G. J. H., Grootjen,

M., van Veenendaal, J., 2003. Assessing Cognitive

Load Distributions for Envisioned Task Allocations

and Support Functions, Thirteenth International Ship

Control Systems Symposium, Orlando, Florida, USA,

7-9.4.03.

Quintanar, L. R, Crowell, C. R, Pryor, J. B. and

Adamopoulos, J., 1982. Human Computer Interaction:

A Preliminary Social Psychological Analysis,

Behaviour Research Methods and Instrumentation

14(2): 210-220

Sweller, J., van Merrienboer, J. J. G. and Paas, F. G. W.

C., 1998. Cognitive Architecture and Instructional

Design, Educational Psychology Review, 10(3),

Plenum.

Wang, D., Kaufman, D. R., Mendonca, E. A., Seol, Y.,

Johnson, S. B., and Cimino, J. J., 2002. The Cognitive

Demands of an Innovative Query User Interface,

Proceedings of the AIMA 202 Annual Symposium,

850-854.

ICEIS 2008 - International Conference on Enterprise Information Systems

188