WHY ANTHROPOMORPHIC USER INTERFACE FEEDBACK CAN BE EFFECTIVE AND PREFERRED BY USERS

: This paper addresses and resolves an interesting question concerning the reason for anthropomorphic user interface feedback being more effective (in two of three contexts) and preferred by users compared to an equivalent non-anthropomorphic feedback. Firstly the paper will summarise the author’s three internationally published experiments and results. These will show statistically significant results indicating that in two of the three contexts anthropomorphic user interface feedback is more effective and preferred by users. Secondly some of the famous work by Reeves and Nass will be introduced. This basically shows that humans behave in a social manner towards computers through a user interface. Thirdly the reasons for the obtained results by the author are inextricably linked to the work of Reeves and Nass. It can be seen that the performance results and preferences are due to the subconscious social behaviour of humans towards computers through a user interface. The conclusions reported in this paper are of significance to user interface designers as they allow one to design interfaces which match more closely our human characteristics. These in turn would enhance the profits of a software house.


INTRODUCTION
User interface feedback in software systems is being improved as time passes and developers dedicate more time to the feedback and realise that feedback to the user is just as important as the rest of an application.
In line with the goal of constant improvement and better understanding of user interface feedback this research has looked at the effectiveness and user approval of anthropomorphic feedback.This was compared to an equivalent non-anthropomorphic feedback.
Anthropomorphism at the user interface usually involves assigning human characteristics or qualities or both to something which is not human, e.g. a talking dog or a cube with a face that can talk etc.A well known example is the Microsoft Office Paper Clip.It could also be the actual manifestation of a real human such as a video of a human (Bengtsson et al, 1999).
This issue has been considered because there was a division between computer scientists where certain computer scientists are against (e.g.chapter by Shneiderman in ((Bradshaw, 1997) and (Shneiderman, 1992)) anthropomorphism at the user interface and others are in favour (e.g.Agarwal (1999), Cole et al. (1999), Dertouzos (1999), Guttag (1999), Koda andMaes (1996a), (1996b), Maes (1994) and Zue (1999)) of using anthropomorphism at the user interface.However there has not been concrete enough evidence to show which opinion may be correct.
Experiments (summarised below and detailed in Murano (2001a(summarised below and detailed in Murano ( ), (2001b(summarised below and detailed in Murano ( ), (2002a(summarised below and detailed in Murano ( ), (2002b(summarised below and detailed in Murano ( ), (2003))) have been conducted where it has been shown with statistical significance that in certain contexts anthropomorphic user interface feedback is more effective and preferred by users.However these experiments concentrated on 'what' type of feedback was better (i.e.anthropomorphic or nonanthropomorphic) and not on 'why' a particular type of feedback was better over the other.This issue of 'why' was raised as an interesting question at various international conferences attended by the author.Hence firstly this paper aims to address this question and provide an answer by means of the body of evidence produced by Reeves and Nass.It is believed by the author that no other researchers outside of Reeves and Nass' influence have used and validated some of their results in such a detailed manner.Secondly, the experiments conducted and summarised below, are innovative in that while they follow the guidelines of Reeves and Nass, this is the first time that the guidelines have been applied to a more realistic context.The experiments by Reeves and Nass were more artificial in nature.This is because despite many computers and applications being in homes and businesses, there are still many prospective users who are afraid of computers.These prospective users could become actual enthusiastic users, thus potentially increasing business profits for a software house.This could be achieved by the improvement of user interface feedback by using the findings of this paper.

SUMMARY OF EXPERIMENTS
In the next three sections below a brief summary is presented of the three experiments.Full details for repeatability can be found in Murano (2001aMurano ( ), (2001bMurano ( ), (2002aMurano ( ), (2002bMurano ( ), (2003)).However for each of the three experiments within users' designs were used.This meant that in each experiment all the subjects tried all tasks and had the opportunity to use all relevant kinds of feedback.Considerable efforts were made to maintain laboratory conditions constant for each subject.Also efforts were made to control possible confounding variables.

Experiment One
The first experiment Murano (2002a) was in the context of software for in-depth understanding.This was specifically English as a foreign language (EFL) pronunciation.The language group used was Italian native speakers who did not have 'perfect' English.Software was specifically designed to automatically handle user speech via an automatic speech recognition (ASR) engine.Further, in line with EFL literature by Kenworthy (1992) and Ur (1996) exercises were designed and incorporated as part of the software to test problem areas that Italian speakers have when pronouncing English.
Anthropomorphic feedback in the form of a video of a real EFL tutor giving feedback was designed.This in effect was a set of dynamically loaded video clips which were activated based on the software's decision concerning the potential error a user had done (if no errors were made no pronunciation corrections were made by the software).This type of feedback was compared against a non-anthropomorphic equivalent.In this case two-dimensional diagrams with guiding text were used.The diagrams were facial cross-sections aiming to assist a user in the positioning of their mouth and tongue etc. for the relevant pronunciation of a given exercise.This type of feedback was based on EFL principles found in Baker (1981) and Baker (1998).No feedback type was ever tied to the same exercise, i.e. feedback was randomly assigned to an exercise.
The results for 18 Italian users (with imperfect English pronunciation) taking part in a tightly controlled experiment, going through a series of exercises were statistically significant.Users were scored (scores used in hypothesis testing statistical analysis) according to the number of attempts they had to make to complete an exercise successfully.
The statistical results suggested the anthropomorphic feedback to be more effective.Users were able to self-correct their pronunciation errors more effectively with the anthropomorphic feedback.The scores obtained were approximately normally distributed.These were then used in a ttest.The results are in the Table 1.
Furthermore it was clear that users preferred the anthropomorphic feedback.The actual scores obtained from the questionnaires using a Likert scale, where 1 was a negative response and 9 was a positive response, are detailed in Table 2.
Hence it was concluded that the statistically significant results suggested the anthropomorphic

Table 1 :
Comparison of Video Vs Diagrams and text Comparison of Video Vs.