USING MBIUI LIFE-CYCLE FRAMEWORK FOR AN

AFFECTIVE BI-MODAL USER INTERFACE

Katerina Kabassi

Department of Ecology and the Environment, Technological Educational Institute of the Ionian Islands

2 Kalvou Sq., 29100 Zakynthos, Greece

Maria Virvou, Efthymios Alepis

Department of Informatics, University of Piraeus

80 Karaoli & Dimitriou St., 18534, Piraeus, Greece

Keywords: Software life-cycle, multi-criteria decision making theories, affective interaction, user modelling.

Abstract: Decision making theories seem very promising

for improving human-computer interaction. However, the

actual process of incorporating multi-criteria analysis into an intelligent user interface involves several

development steps that are not trivial. Therefore, we have employed and tested the effectiveness of a

unifying life-cycle framework that may be used for the application of many different multi-criteria decision

making theories. The life-cycle framework is called MBIUI and in this paper we show how we have used it

for employing a multi-criteria decision making theory, called Simple Additive Weighting, in an affective bi-

modal educational system. More specifically, we describe the experimental studies for designing,

implementing and testing the decision making theory. The decision making theory has been adapted in the

user interface for combining evidence from two different modes and providing affective interaction.

1 INTRODUCTION

Decision making theories are rather promising for

knowledge-based software. However, multi-criteria

analysis has not been used adequately in intelligent

user interfaces, even though user-computer

interaction is, by nature, multi-criteria-based. The

actual process of incorporating multi-criteria

analysis into an intelligent user interface is neither

clearly defined nor adequately described in the

literature. Furthermore, Hull et al. (2002) point out

that as the systems become more complex, their

development and maintenance is becoming a major

challenge. This is particularly the case for software

that incorporates intelligence. Indeed, intelligent

systems are quite complex and they have to be

developed based on software engineering

approaches that are quite generic and do not

specialise on the particular difficulties of the

intelligent approach that is to be used.

For this purpose a life-cycle framework have

been

developed for the incorporation of a multi-

criteria theory in an Intelligent User Interface (IUI).

This framework is called MBIUI (Multi-criteria

Based Intelligent User Interface) life-cycle

framework (Kabassi & Virvou 2006) and involves

the description of a software life-cycle that gives

detailed information and guidelines about the

experiments that need to be conducted, the design of

the software, the selection of the right decision

making theory and the evaluation of the IUI that

incorporates a decision making theory.

MBIUI life-cycle framework is based on the

ational Unified Process (RUP) (Jacobson et al.

1999). RUP is clearly documented and easily used

due to its clarity. For this purpose RUP seems quite

suitable for the development of knowledge-based

systems. Indeed, RUP’s rationale has also been

adapted in other methodologies for knowledge-based

systems, e.g. ADELFE (Bernon et al. 2003).

In this paper, we test the effectiveness of this

fram

ework by applying it for the development of an

affective bi-modal user interface. The particular

interface is called Edu-Affe-Mikey and is an

affective educational user interface targeted to first-

year medical students. Emphasis has been given on

the application of the MBIUI life-cycle framework

for the application of a multi-criteria decision

Kabassi K., Virvou M. and Alepis E. (2007).

USING MBIUI LIFE-CYCLE FRAMEWORK FOR AN AFFECTIVE BI-MODAL USER INTERFACE.

In Proceedings of the Second International Conference on Software and Data Technologies - SE, pages 40-47

 SciTePress

making method for combining evidence from two

different modes in order to identify the users’

emotions. More specifically, the Simple Additive

Weighting (SAW) (Fishburn 1967, Hwang & Yoon

1981) has been applied in the educational user

interface for evaluating different emotions, taking

into account the input of the two different modes and

selecting the one that seems more likely to have

been felt by the user. In this respect, emotion

recognition is based on several criteria that a human

tutor would have used in order to perform emotion

recognition of his/her students during the teaching

course.

A main difference of the proposed approach with

other systems that employ decision making theories

(Naumann 1998; Schütz & Schäfer 2001;

Bohnenberger et al. 2005; Kudenko et al. 2003) is

that the values of the weights of the criteria are not

static. More specifically, the values of the criteria

used in the proposed approach are acquired by user

stereotypes and differ for the different categories of

users. Stereotypes constitute a common user

modelling technique for drawing assumptions about

users belonging to different groups (Rich 1989;

1999). In our case, user stereotypes have been

constructed with respect to the different emotional

states of users that are likely to occur in typical

situations during the educational process and their

interaction with the educational software and

represent the weight of the criteria.

The main body of this paper is organized as

follows: In section 2 we present and discuss the

MBIUI life-cycle framework. In sections 3 and 4 we

present briefly the experimental studies for

requirements capture and analysis. Section 5

describes the design of the affective bi-modal

educational application and section 6 its main

characteristics. In section 7 we present and discuss

the results of the evaluation of the multi-criteria

model. Finally, in section 8 we give the conclusions

drawn from this work.

2 MBIUI LIFE-CYCLE

FRAMEWORK

MBIUI life-cycle framework is based on RUP,

which gives a framework of a software life-cycle

that is based on iterations and maintains its phases

and procedural steps. However, RUP does neither

specify what sort of requirements analysis has to

take place nor what kind of prototype has to be

produced during each phase or procedural step. Such

specifications are provided by our MBIUI

framework concerning IUIs that are based on multi-

criteria theories.

According to MBIUI framework, during the

inception phase, the requirements capture is

conducted. During requirements capture, a prototype

is developed and the main requirements of the user

interface are specified. At this point the multi-

criteria decision making theory that seems most

promising for the particular application has to be

selected. This decision may be revised in the

procedural step of requirements capture in the phase

of construction.

According to MBIUI, in the inception phase,

during analysis, two different experiments are

conducted in order to select the criteria that are used

in the reasoning process of the human advisors as

well as their weights of importance. The

experiments should be carefully designed, since the

kind of participants as well as the methods selected

could eventually affect the whole design of the IUI.

Both experiments involve human experts in the

domain being reviewed.

The information collected during the two

experiments of the empirical study is further used

during the design phase of the system, where the

decision making theory that has been selected is

applied to the user interface. More specifically, in

the elaboration phase, during design, the user

modelling component of the system is designed and

the decision making model is adapted for the

purposes of the particular domain. Kass and Finin

(1989) define the user model as the knowledge

source of a system that contains hypotheses

concerning the user that may be important in terms

of the interactive behaviour of the system.

In the elaboration phase, during implementation,

the user modelling component of the system as well

as the basic decision making mechanisms are

developed. As a result a new version of the IUI is

developed which incorporates fully the multi criteria

decision making theory.

In the construction phase, during testing, the IUI

that incorporates the multi-criteria decision making

theory is evaluated. The evaluation of IUIs is very

important for their accuracy, efficiency and

usefulness. In MBIUI, evaluation is considered

important for two reasons: 1) the effectiveness of the

particular decision making theory that has been used

has to be evaluated 2) the effectiveness of the IUI in

general has to be evaluated. In case the version of

the IUI that incorporates a particular decision

making theory does not render satisfactory

evaluation results with respect to real users and

human experts, then the designers have to return to

requirements capture, select an alternative decision

making model and a new iteration of the life cycle

takes place. In transition phase, during testing, the

decision making model that has been finally selected

USING MBIUI LIFE-CYCLE FRAMEWORK FOR AN AFFECTIVE BI-MODAL USER INTERFACE

is evaluated and possible refinements of that model

may take place, if this is considered necessary.

3 REQUIREMENTS CAPTURE

In the inception phase, during the procedural step of

requirements capture the basic requirements of the

system are specified. For this purpose we conducted

an empirical study. Due to the difference of an

affective bi-modal user interface from common user

interfaces, the main aim of the particular experiment

was to find out how users express their emotions

through a bi-modal interface that combines voice

recognition and input from keyboard.

50 users (male and female), of the age range 17-

19 and at the novice level of computer experience

participated the experiment. The particular users

were selected because such a profile describes the

majority of first year medical students in a Greek

university that the educational application is targeted

to. They are usually between the age of 17 and 19

and usually have only limited computing experience,

since the background knowledge required for

medical studies does not include advanced computer

skills.

These users were given questionnaires

concerning their emotional reactions to several

situations of computer use in terms of their actions

using the keyboard and what they say. Participants

were asked to determine what their possible

reactions may be when they are at certain emotional

states during their interaction. Our aim was to

recognise the possible changes in the users’

behaviour and then to associate these changes with

emotional states like anger, happiness, boredom, etc.

After collecting and processing the information of

the empirical study we came up with results that led

to the design of the affective module of the

educational application. For this purpose, some

common positive and negative feelings were

identified.

The results of the empirical study were also used

for designing the user stereotypes. In our study user

stereotypes where built first by categorizing users by

their age, their educational level and by their

computer knowledge level. The underlying

reasoning for this is that people’s behaviour while

doing something may be affected by several factors

concerning their personality, age, experience, etc.

For example, experienced computer users may be

less frustrated than novice users or older people may

have different approaches in interacting with

computers, comparing with younger people.

Younger computer users are usually more expressive

than older users while interacting with an animated

agent and we may expect to have more data from

audio mode than by the use of a keyboard. The same

case is when a user is less experienced in using a

computer than a user with a high computer

knowledge level. In all these cases stereotypes were

used to indicate which specific characteristics in a

user’s behaviour should be taken to account in order

make more accurate assumptions about the users’

emotional state.

4 ANALYSIS

According to MBIUI, during analysis, two different

experiments are conducted. The first experiment

aims at determining the criteria that are used in the

reasoning process of the human advisors and the

second aims at calculating their weights of

importance.

4.1 Determining Multiple Criteria

Decision making theories provide precise

mathematical methods for combining criteria in

order to make decisions but do not define the

criteria. Therefore, in order to locate the criteria that

human experts take into account while providing

individualised advice, we conducted an empirical

study.

The empirical study should involve a satisfactory

number of human experts, who will act as the human

decision makers and are reviewed about the criteria

that they take into account when providing

individualised advice. Therefore, in the experiment

conducted for the application of the multi-criteria

theory in the e-learning system, 16 human experts

were selected in order to participate in the empirical

study. All the human experts possessed a first and/or

higher degree in Computer Science.

The participants of the empirical study were

asked which input action from the keyboard and the

microphone would help them find out what the e-

motions of the users were. From the input actions

that appeared in the experiment, only those proposed

by the majority of the human experts were selected.

In particular considering the keyboard we have: a)

user types normally b) user types quickly (speed

higher than the usual speed of the particular user) c)

user types slowly (speed lower than the usual speed

of the particular user) d) user uses the backspace key

often e) user hit unrelated keys on the keyboard f)

user does not use the keyboard.

Considering the users’ basic input actions

through the microphone we have 7 cases: a) user

speaks using strong language b) users uses

ICSOFT 2007 - International Conference on Software and Data Technologies

exclamations c) user speaks with a high voice

volume (higher than the average recorded level) d)

user speaks with a low voice volume (low than the

average recorded level) e) user speaks in a normal

voice volume f) user speaks words from a specific

list of words showing an emotion g) user does not

say anything.

4.2 Determining the Weights of

Importance of the Criteria

During requirements capture the main categories of

the users and emotions were identified. As a result

the main stereotypes were designed. For the design

of the body of the stereotype we have used the

results of the empirical study described in section

4.1, in we categorized users’ input actions in terms

of the two modes of the bi-modal system. These

actions would indicate possible changes in a user’s

emotional state while s/he interacted with the

system. However, in order to identify the weights of

the criteria (input action) another experimental study

was conducted.

More specifically, 50 medical students were

asked to use Edu-Affe-Mickey, which incorporated a

user modelling component. The user modelling

component recorded all users’ actions as a filter

between the user and the main educational

application. Then these actions were classified by

the six and seven basic input actions in regard to the

keyboard and the microphone respectively.

Table 1: Values for the stereotypic weights for the

emotions of happiness and anger concerning input from

the keyboard.

Using the keyboard

Emotion of

happiness

Emotion of anger

input

action

weight

input

action

weight

k1 0,4 k1 0,11

k2 0,4 k2 0,14

k3 0,1 k3 0,18

k4 0,05 k4 0,2

k5 0,05 k5 0,25

k6 0 k6 0,12

The results of the empirical study were collected

and analyzed. The analysis revealed how important

each input action is in identifying each emotion.

Therefore, the weight of each criterion (input action)

for all emotions were identified and the default

assumptions of the stereotypes were designed.

Tables 1 and 2 illustrate the values of the

weights for two opposite (positive/negative)

emotions, namely the emotion of happiness and the

emotion of anger. Variables k1 to k6 refer to the

weights of the six basic input actions from the

keyboard, while variables m1 to m7 refer to the

weights of the seven possible input cases concerning

interaction through the microphone. We may also

note that for each emotion and for each mode the

values of the weights have sum that equals to 1.

Table 2: Values for the stereotypic weights for the

emotions of happiness and anger concerning input from

the microphone.

Using the microphone

Emotion of

happiness

Emotion of anger

input

action

weight

input

action

weight

m1 0,06 m1 0,19

m2 0,18 m2 0,09

m3 0,15 m3 0,12

m4 0,02 m4 0,05

m5 0,14 m5 0,12

m6 0,3 m6 0,27

m7 0,15 m7 0,16

5 DESIGN

In MBIUI the design of the running application is

mainly concerned with the design of the user model

and is divided into two major parts with respect to

the application of a multi-criteria decision making

theory: 1) design decisions about how the values of

the criteria are estimated based on the information of

the user model and 2) design of the embedment of

the actual multi-criteria theory that has been selected

into the system.

The input actions that were identified by the

human experts during the first experimental study of

analysis provided information for the actions that

affect the emotional states that may occur while a

user interacts with an educational system. These

input actions are considered as criteria for evaluating

all different emotions and selecting the one that

seems more prevailing. More specifically, each

emotion is evaluated first using only the criteria

(input actions) from the keyboard and then only the

criteria (input actions) from the microphone. For the

evaluation of each alternative emotion the system

uses SAW for a particular category of users.

According to SAW, the multi-criteria utility function

USING MBIUI LIFE-CYCLE FRAMEWORK FOR AN AFFECTIVE BI-MODAL USER INTERFACE

for each emotion in each mode is estimated as a

linear combination of the values of the criteria that

correspond to that mode.

In view of the above, for the evaluation of each

emotion taking into account the information

provided by the keyboard is done using formula 1.

(Formula 1)

Similarly, for the evaluation of each emotion

taking into account the information provided by the

other mode (microphone) is done using formula 2.

(Formula 2)

is the probability that an emotion has

occurred based on the keyboard actions and

is the probability that refers to an emotional state

using the users’ input from the microphone.

and take their values in [0,1].

In formula 1 the k’s from k1 to k6 refer to the six

basic input actions that correspond to the keyboard.

In formula 2 the m’s from m1 to m7 refer to the

seven basic input actions that correspond to the

microphone. These variables are Boolean. In each

moment the system takes data from the bi-modal

interface and translates them in terms of keyboard

and microphone actions. If an action has occurred

the corresponding criterion takes the value 1,

otherwise its value is set to 0. The w’s represent the

weights. These weights correspond to a specific

emotion and to a specific input action and are

acquired by the stereotype database. More

specifically, the weights are acquired by the

stereotypes about the emotions.

In cases where both modals (keyboard and

microphone) indicate the same emotion then the

probability that this emotion has occurred increases

significantly. Otherwise, the mean of the values that

have occurred by the evaluation of each emotion

using formulae 1 and 2 is calculated.

The system compares the values from all the

different emotions and determines whether an

emotion is taking effect during the interaction.

As an example we give the two formulae with

their weights for the two modes of interaction that

correspond to the emotion of happiness when a user

(under the age of 19 and novice with respect to

his/her computer skills) gives the correct answer in a

test of our educational application. In case of

considering the keyboard we have:

65432111

005.005.01.04.04.0

kkkkkkem

+++++=

44133122111111

11111

kwkwkwkwem

kekekekee

+++=

In this formula, which corresponds to the

emotion of happiness, we can observe that the

highest weight value correspond to the normal and

quickly way of typing. Slow typing, ‘often use of the

backspace key’ and ‘use of unrelated keys’ are

actions with lower values of stereotypic weights.

Absence of typing is unlikely to take place.

Concerning the second mode (microphone) we have:

661551

kwkw

keke

765432121

15.03.014.002.015.018.006.0

mmmmmmmem

+++

44133122111121

11111

mwmwmwmwem

memememee

+++=

In the second formula, which also corresponds to

the emotion of happiness, we can see that the highest

weight corresponds to m6, which refers to the

‘speaking of a word from a specific list of words

showing an emotion’ action. The empirical study

gave us strong evidence for a specific list of words.

In the case of words that express happiness, these

words are more likely to occur in a situation where a

novice young user gives a correct answer to the

system. Quite high are also the weights for variables

m2 and m3 that correspond to the use of

exclamations by the user and to the raising of the

user’s voice volume. In our example the user may do

something orally or by using the keyboard or by a

combination of the two modes. The absence or

presence of an action in both modes will give the

Boolean values to the variables k1…k6 and m1…m7.

771661551

111

mwmwmw

mememe

A possible situation where a user would use both

the keyboard and the microphone could be the

following: The specific user knows the correct

answer and types in a speed higher than the normal

speed of writing. The system confirms that the

answer is correct and the user says a word like

‘bravo’ that is included in the specific list of the

system for the emotion of happiness. The user also

speaks in a higher voice volume. In that case the

variables k1, m3 and m6 take the value 1 and all the

others are zeroed. The above formulas then give us

and

4.01*4.0

45.01*3.01*15.0

In the same way the system then calculates the

corresponding values for all the other emotions

using other formulae. For each basic action in the

educational application and for each emotion the

corresponding formula have different weights

deriving from the stereotypical analysis of the

empirical study. In our example in the final

comparison of the values for the six basic emotions

the system will accept the emotion of happiness as

the most probable to occur.

2111

emem +

ICSOFT 2007 - International Conference on Software and Data Technologies

6 IMPLEMENTATION

During implementation, in the elaboration phase the

overall functionality and emotion recognition

features of Edu-Affe-Mikey are implemented. The

architecture of Edu-Affe-Mikey consists of the main

educational application with the presentation of

theory and tests, a programmable human-like

animated agent, a monitoring user modeling

component and a database.

Figure 1: A screen-shot of theory presentation in Edu-

Affe-Mikey educational application.

While using the educational application from a

desktop computer, students are being taught a

particular medical course. The information is given

in text form while at the same time the animated

agent reads it out loud using a speech engine. The

student can choose a specific part of the human body

and all the available information is retrieved from

the systems’ database. In particular, the main

application is installed either on a public computer

where all students have access, or alternatively each

student may have a copy on his/her own personal

computer. An example of using the main application

is illustrated in figure 1. The animated agent is

present in these modes to make the interaction more

human-like.

While the users interact with the main

educational application and for the needs of emotion

recognition a monitoring component records the

actions of users from the keyboard and the

microphone. These actions are then processed in

conjunction with the multi-criteria model and

interpreted in terms of emotions. The basic function

of the monitoring component is to capture all the

data inserted by the user either orally or by using the

keyboard and the mouse of the computer. The data is

recorded to a database and the results are returned to

the basic application the user interacts with. Figure 2

illustrates the “monitoring” component that records

the user’s input and the exact time of each event.

Figure 2: Snapshot of operation of the user modeling

component.

Instructors have also the ability to manipulate the

agents’ behaviour with regard to the agents’ on

screen movements and gestures, as well as speech

attributes such as speed, volume and pitch.

Instructors may programmatically interfere to the

agent’s behaviour and the agent’s reactions

regarding the agents’ approval or disapproval of a

user’s specific actions. This adaptation aims at

enhancing the “affectiveness” of the whole

interaction. Therefore, the system is enriched with

an agent capable to express emotions and, as a

result, enforces the user’s temper to interact with

more noticeable evidence in his/her behaviour.

7 TESTING

In construction phase, during the procedural step of

testing, the final version of the system is evaluated.

When a user interface incorporates a decision

making theory, the evaluation phase plays an

important role for showing whether the particular

theory is effective or not. In MBIUI life-cycle

framework it is considered important to conduct the

evaluation of a decision making model by

comparing the IUI’s reasoning with that of real

users. Therefore, in this experiment it is important to

evaluate how successful the application of the

decision making model is in selecting the alternative

action that the human experts would propose in the

case of a user’s error. For this reason, it has to be

checked whether the alternative actions that are

proposed by the human experts are also highly

ranked by the application of the decision making

USING MBIUI LIFE-CYCLE FRAMEWORK FOR AN AFFECTIVE BI-MODAL USER INTERFACE

model. In case this comparison reveals that the

decision making model is not adequate, another

iteration of the life-cycle has to take place and

another decision model should be selected. This

iteration continues until the evaluation phase gives

satisfactory results.

In view of the above, an evaluation study was

conducted. Therefore, the 50 medical students that

were involved in the empirical study during

requirements capture were also involved in the

evaluation of the multi-criteria emotion recognition

system. More specifically, they were asked to

interact with the educational software and the whole

interaction was video recorded. The protocols

collected were presented to the same users in order

to perform emotion recognition for themselves with

regard to the six emotional states, namely happiness,

sadness, surprise, anger, disgust and the neutral

emotional state.

The students as observers were asked to justify

the recognition of an emotion by indicating the

criteria that s/he had used in terms of the audio mode

and keyboard actions. Whenever a participant

recognized an emotional state, the emotion was

marked and stored as data in the system’s database.

Finally, after the completion of the empirical study,

the data were compared with the systems’

corresponding hypothesis in each case an emotion

was detected. Table 3 illustrates the percentages of

successful emotion recognition of each mode after

the incorporation of stereotypic weights and the

combination through the multi-criteria approach.

Table 3. Recognition of emotions using stereotypes and

SAW theory.

Using Stereotypes and SAW

Emotions

Audio

mode

recognition

Recognition

through

keyboard

Multi-

criteria bi-

modal

recognition

Neutral 17% 32% 46%

Happiness 52% 39% 64%

Sadness 65% 34% 70%

Surprise 44% 8% 45%

Anger 68% 42% 70%

Disgust 61% 12% 58%

Provided the correct emotions for each situation

identified by the user himself/herself we were able

to come up with conclusions about the efficiency of

our systems’ emotion recognition ability. However,

one may notice that there are a few cases where the

proposed approach had a bit worse performance in

recognizing an emotional state (e.g. neutral

emotional state). Possible reason is the fact that

some emotional states, e.g. neutral emotional state,

give little evidence to certain modes (e.g. the

keyboard mode). For the same emotions, other

modalities (e.g. the visual mode) may give us

significantly better evidence. However, the

combination of the two modes in the multi-criteria

model increases significantly the accuracy of the

proposed approach.

Although, success rates may look at a first

glance lower than expected, we should underline the

fact that emotion recognition is a very difficult task

for human and their success rates are also quite low.

Therefore, the results of the evaluation study offer

evidence for the adequacy of the multi-criteria multi-

modal model for emotion recognition.

8 CONCLUSIONS

In this paper, we have used a general framework,

which provides detailed guidelines for the

application of a multi-criteria decision making

theory in an affective bi-modal Intelligent User

Interface. This framework is called MBIUI life-cycle

framework and was initially designed for

incorporating a decision making theory in a user

interface that helps users during their interaction

with a file-store system.

In this paper we aimed at checking the

effectiveness of this framework in order to apply a

simple multi-criteria decision making theory in an

adaptive bi-modal user interface. Indeed, the MBIUI

life-cycle framework facilitates the application of

the multi-criteria decision making theory by

providing detailed guidelines for all experimental

studies during requirements capture and analysis as

well as testing.

The user interface presented in this paper is used

for providing medical education to first-year medical

students and is called Edu-Affe-Mikey. In this

system, the multi-criteria decision making theory,

SAW, is used for combining evidence from two

different modes in order to identify the users’

emotions. SAW is used for evaluating different

emotions, taking into account the input of the two

different modes and selecting the one that seems

more likely to have been felt by the user. The

particular user interface offers affective bi-modal

interaction and for this reason differentiates from the

other user interfaces. The fact that the particular

framework can be used in interfaces that differ in

many ways strengthens MBIUI’s generality.

The affective bi-modal user interface has been

evaluated and the results prove the effectiveness of

the multi-criteria decision making theory for

ICSOFT 2007 - International Conference on Software and Data Technologies

combining evidence from two different modes and

perform emotion recognition.

ACKNOWLEDGEMENTS

Support for this work was provided by the General

Secretariat of Research and Technology, Greece,

under the auspices of the PENED-2003 program.

REFERENCES

Bernon, C., Gleizes, M.P., Peyruqueou, S., Picard, G.,

2003. ADELFE: A methodology for adaptive multi-

agent systems engineering, Engineering Societies in

the Agents World III, Lecture Notes in Artificial

Intelligence, Vol. 2577, pp. 156-169.

Bohnenberger, T. , Jacobs, O., Jameson A., Aslan, I. 2005.

Decision-Theoretic Planning Meets User

Requirements: Enhancements and Studies of an

Intelligent Shopping Guide, in H. Gellersen, R. Want,

& A. Schmidt (Eds.), Pervasive computing: Third

international conference, Berlin: Springer, 279–296.

Fishburn, P.C. 1967. Additive Utilities with Incomplete

Product Set: Applications to Priorities and

Assignments, Operations Research.

Hull, M. E. C. , Taylor, P. S., Hanna, J. R. P., Millar, R. J.

2002. Software development processes- an

assessment, Information and Software Technology,

vol. 44, pp. 1-12.

Hwang C.L., Yoon, K., 1981. Multiple Attribute Decision

Making: Methods and Applications, Lecture Notes in

Economics and Mathematical Systems, vol. 186.

Jacobson, I., Booch, G., Rumbaugh, J. 1999. The Unified

Software Development Process, Addison-Wesley,

Reading, MA, 1999.

Kabassi, K., Virvou, M., 2006. A Knowledge-based

Software Life-Cycle Framework for the incorporation

of Multi-Criteria Analysis in Intelligent User

Interfaces. IEEE Transactions on Knowledge and

Data Engineering, vol. 18, No. 9, pp. 1-13.

Kass, R., Finin, T. 1989. The role of User Models in

Cooperative Interactive Systems, International

Journal of Intelligent Systems, vol. 4, pp. 81-112.

Kudenko, D., Bauer, M. Dengler, D. 2003. Group

Decision Making Through Mediated Discussions,

Proceedings of the 9th International Conference on

User Modelling, 2003.

Naumann, F., 1998. Data Fusion and Data Quality,

Proceedings of the New Techniques and Technologies

for Statistics.

Rich, E., 1989. Stereotypes and User Modeling. In: Kobsa,

A., Wahlster, W., (Eds.) User Models in Dialog

Systems, pp. 199-214.

Rich, E., 1999. Users are individuals: individualizing user

models. International Journal of Human-Computer

Studies vol. 51, pp. 323-338.

Schütz, W., Schäfer, R., 2001. Bayesian networks for

estimating the user's interests in the context of a

configuration task, in R. Schäfer, M. E. Müller, and S.

A. Macskassy (eds.) Proceedings of the UM2001

Workshop on Machine Learning for User Modeling,

pp. 23-36.

USING MBIUI LIFE-CYCLE FRAMEWORK FOR AN AFFECTIVE BI-MODAL USER INTERFACE