user modelling in ambient intelligence for elderly and disabled people

12
User Modelling in Ambient Intelligence for Elderly and Disabled People Roberto Casas 1 , Rub´ en Blasco Mar´ ın 1 Alexia Robinet 2 , Armando Roy Delgado 2 , Armando Roy Yarza 1 , John McGinn 2 , Richard Picking 2 , and Vic Grout 2 1 Grupo Tecnodiscap, Universidad de Zaragoza, Maria de Luna 1 50018 Zaragoza, Spain (rcasas|rblasco|armanroy)@unizar.es, 2 Centre for Applied Internet Research (CAIR), University of Wales, NEWI, Plas Coch Campus, Mold Road, Wrexham, LL11 2AW, Wales, UK (a.robinet|a.delgado|j.mcginn|r.picking|v.grout)@newi.ac.uk Abstract. Ambient Intelligence (AmI) characterizes a vision where hu- mans are surrounded by computers. Combining ongoing technological de- velopments (e.g. pervasive computing, wearable devices, sensor networks etc.) with user-centred design methods greatly increases the acceptance of the intelligent system and makes it more capable of providing a better quality of life in a non-intrusive way. Elderly people, with or without disabilities, could clearly benefit from this concept. Thanks to smart environments, they can experience considerable enhancements, giving them an opportunity to live more independently and for longer in their home rather than in a health-care centre. However, to implement such a system, it is essential to know for whom we are designing. Indeed, the system needs to know the users’ capabilities and behaviour in order to adapt itself and improve the interaction. Creating a user model based on the principal characteristics of the end-users will contribute positively in the development of the system’s intelligence. In this paper, we present an intelligent system with a monitoring infrastructure that will help mainly elderly users with impairments to overcome their handicap. The purpose of such a system is to create a safe and intuitive environment that will facilitate the achievement of household tasks in order to preserve inde- pendence of elderly residents for a while longer. Pursuing this goal, we propose to use the persona concept to help us build a user model based on the personas’ aptitudes. The practice of user modelling emphasizes the importance of user-centred techniques in any AmI system develop- ment and highlights the potential impacts of AmI for certain targeted groups - in this case, the elderly and people with disabilities. Key words: User models, Ambient Intelligence, elderly people, assistive interfaces, impairments

Upload: tqg

Post on 27-Apr-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

User Modelling in Ambient Intelligence forElderly and Disabled People

Roberto Casas1, Ruben Blasco Marın1 Alexia Robinet2, Armando RoyDelgado2, Armando Roy Yarza1, John McGinn2, Richard Picking2, and Vic

Grout2

1 Grupo Tecnodiscap, Universidad de Zaragoza, Maria de Luna 150018 Zaragoza, Spain

(rcasas|rblasco|armanroy)@unizar.es,2 Centre for Applied Internet Research (CAIR), University of Wales, NEWI, Plas

Coch Campus, Mold Road,Wrexham, LL11 2AW, Wales, UK

(a.robinet|a.delgado|j.mcginn|r.picking|v.grout)@newi.ac.uk

Abstract. Ambient Intelligence (AmI) characterizes a vision where hu-mans are surrounded by computers. Combining ongoing technological de-velopments (e.g. pervasive computing, wearable devices, sensor networksetc.) with user-centred design methods greatly increases the acceptanceof the intelligent system and makes it more capable of providing a betterquality of life in a non-intrusive way. Elderly people, with or withoutdisabilities, could clearly benefit from this concept. Thanks to smartenvironments, they can experience considerable enhancements, givingthem an opportunity to live more independently and for longer in theirhome rather than in a health-care centre. However, to implement sucha system, it is essential to know for whom we are designing. Indeed, thesystem needs to know the users’ capabilities and behaviour in order toadapt itself and improve the interaction. Creating a user model based onthe principal characteristics of the end-users will contribute positively inthe development of the system’s intelligence. In this paper, we present anintelligent system with a monitoring infrastructure that will help mainlyelderly users with impairments to overcome their handicap. The purposeof such a system is to create a safe and intuitive environment that willfacilitate the achievement of household tasks in order to preserve inde-pendence of elderly residents for a while longer. Pursuing this goal, wepropose to use the persona concept to help us build a user model basedon the personas’ aptitudes. The practice of user modelling emphasizesthe importance of user-centred techniques in any AmI system develop-ment and highlights the potential impacts of AmI for certain targetedgroups - in this case, the elderly and people with disabilities.

Key words: User models, Ambient Intelligence, elderly people, assistiveinterfaces, impairments

2

1 Introduction

With the major increase of the elderly population, particularly in Europe, thereis a need for smart homes. In fact, it seems that keeping the elderly out of in-stitutional care is the main concern in order to ease the demand made uponhealth services [1]. In the context of the ageing population, Ambient Intelligence(AmI) offers solutions such as health monitoring, location tracking, security etc.that could appeal to many people if only cost was not a barrier. Financially, liv-ing in a smart home is preferable as living in a nursing home is quite expensive.Smart home environments can perceive long-term changes that may cause healthconcerns [2]. Such a system, embedded in smart homes, could alert carers andfamily of any significant changes in the resident’s behaviour, diet, daily tasks orhealth. Fall detectors, smart pill dispensers, RFID tags on food packaging, med-ical equipment to test heart rate and blood pressure, GPS trackers, sensors, andso on, create a safer environment in which to live for people with sensory, cog-nitive or physical challenges. This convergence of technologies enables elderlypeople to stay at home and receive human care in a much quicker and easiermanner.

The application of the information society to home environments is charac-terized by the integration of networked computational devices into the physicalcontext. This trend, associated with ubiquitous computing, is evolving towardsystems with intelligent and context-sensitive behaviour - a vision of future tech-nological development in AmI [3]. AmI is characterized by ubiquitous computing,omnipresent communications and intelligent user interfaces. AmI systems haveto be adaptive, personalized, unobtrusive and anticipatory [4][5]. As a systemdevelopment trend, universal access to AmI environments brings about the ac-cessibility and usability by user with different characteristics and needs [4]. It iswidely accepted that AmI and ubiquitous computing can cope with the elderlyand people with disabilities’ problems in their everyday life [6][7][8]. For homeenvironment systems, a particular feature is to support daily routines, such aspreparing food or operating household appliances. A typical schema of an AmIapplication is shown in Figure 1.

3

Fig. 1. Example of an AmI architecture

Once the contextual information is collected, there is a data fusion processwhose output will be evaluated. This processed information will be used as inputto the decision process, applying action rules previously set. As a result of the lastprocess, a variety of actions will be performed. For example, ”smart” appliancescan learn a user’s habits, keep track of planned activities and assist in theircompletion [7].

2 User Modelling

An AmI system has to be adaptive (its behaviour can change in response to a per-son’s actions and environment) and personalized (its behaviour can be tailoredto the user’s needs). According to these features, AmI systems for the elderlyand people with disabilities have to adapt, not only to the user’s actions andenvironment, but also to their behaviour and frame of mind. A context-sensitiveAmI system should reconfigure dynamically to accommodate the needs of users,taking into account a wide range of users and context or behaviour situations.This user-centred functioning of AmI systems has to be supported by an ade-quate user model. The intelligence and interface of the system have to be awareof the user abilities and limitations to interact with the person properly. Theuser model must include information about the person’s cognitive level, senso-rial and physical disabilities. The immediate option could be achieved through avery meticulous model, fully parameterizing the person; however, this could bedifficult to use in practice.

4

2.1 Personas

The ”Personas” concept was originally introduced by [9] in his publication ”Theinmates are running the asylum”. In this book, his definition of personas isquoted as follows: ”Personas are not real people, but they represent them through-out the design process. They are hypothetical archetypes of actual users.” [9]There are two different types of personas: primary personas, which representthe main target group and secondary personas, which can use the primary per-sonas’ interfaces but which have specific additional requirements [10][11]. Eventhough personas are fictional characters, they need to be created with rigor andprecision; they tell stories about potential users in ways that allow designers tounderstand them and what they really want. Characteristics such as name, age,picture, profession or any other relevant information are given to each personain order to make them look more realistic or ”alive”. The most accurate way ofcreating personas, also called ”cast of characters”, is to go through a phase ofobservation of real users within the environment in which the system will existand eventually interview them with the intention of finding a common set ofmotivations, behaviours and goals among the end-users. However, this methodis expensive and time-consuming. A low-cost approach is to create them basedon Norman’s assumption personas [12] where designers use their own experienceto identify the different user groups. [11] explain that these assumption-basedpersonas help designers to be aware of legitimate information that can have aneffect on the system’s design.

We have defined ten data-driven personas, based on European statistics(taken from [13] and [14], which both provide key Figures on Europe). Age,education, work, family situation, impairments, technology background and soon were randomly assigned to each personas based on these EU statistics. Alto-gether, we have built the ten personas (of different ages, from different countriesand considering European indicators) presented as follows:

Hannah, 67 years old, Sweden James, 69 years old, United-KingdomFrancesca, 63 years old, Italy Nikos, 62 years old, GreeceEmilie, 83 years old, Belgium Joanna, 76 years old, PolandMikkel, 73 years old, Denmark Manolo, 60 years old, SpainKatharina, 65 years old, Germany Juliette, 70 years old, France

Figure 2 is an example of a persona created with the purpose of defining acommon user profile.

5

Fig. 2. A sample of a Persona

In summary, personas are a valuable tool, particularly when used in scenarioswhere designers test and evaluate the system features for usability and effective-ness. Working with personas is one of the best ways to provide the developerswith valuable insights and an efficient way of keeping the stakeholders in mindthroughout the system design with the aim of making and simplifying design de-cisions. They ”allow us to see the scope and nature of the design problem. Theymake it clear exactly what the user’s goals are, so we can see what the productmust do” [9]

2.2 User modelling method for adaptable systems

User modelling is a field of many years of experience [15] [16] [17] [18]. Over morethan twenty years, researchers have developed different techniques [19] [20] [21]to apply user modelling for both generic and specific purposes in user-centredsystem developments. In the case of smart homes, the user’s acceptance has be-come one of the key factors to determine the success of the system. If the homesystem aims to be universally usable, it will have to accommodate a diverseset of users [22] and adjust to fulfil their needs in case they change. With theaim of helping practitioners to improve their user modelling techniques, someresearches have established rules to follow, as for example, the set of user mod-elling guidelines for adaptive interfaces created by [23].

Thanks to the latest advancements in wireless sensor networks, context-awareness has become an affordable reality for many different applications. Theability to sense the behaviour of the user at home gives the ability to react to

6

changes that clash with the default values set up initially; for example if the sys-tem detects an increase in the user level of expertise it can adjust the interfaceto optimize it with more advanced features and if the user does not respond asexpected it can lower it back to the previous level. When the system is capableof adapting in time, depending on the user needs, applying a user modellingapproach becomes more valuable; it is recommendable not to fall for a designbased on static user stereotypes, which could appear useful in the first instancebut may fail if applied for a long period of time. As it is difficult to develop aninterface that will satisfy every single user, we decided to apply a user modellingtechnique, based on personas, with the intention of creating an accurate, param-eterized user profile that could be adjusted to resolve the User Interface (UI)features of what could be the most appropriate for a specific user at any time.

Building a robust user profile, which all the possible end-users can fit into,including all the relevant user’s characteristics, will help to find the most feasiblesolution from both the point of view of the system’s intelligence and the Human-Machine Interface (HMI). The user modelling process consists of three steps:

1. Sampling: Originally we used the European population, specifically the el-derly and impaired as the overall audience. Following the public Europeanstatistics [13] from the EC we grouped the audience into different groupshaving similar characteristics (e.g. people of certain age, studies, acquisitivepower, disabilities, etc).

2. Analysing: After identifying the relevant characteristics among all the datagathered. Based on probabilistic values, we apply randomly the relevant char-acteristics from each group into ten Personas.

3. Modelling: Finally we used the created Personas to define a practical user’sprofile to be used by the AI of the system - to create user models and mod-ify them if necessary and the HMI - to determine which features the UI willhave on each user’s model.

Figure 3 summarizes the method used to generate the user profile.

7

Fig. 3. The user modelling method

2.3 User profile proposed

The ten personas created represent a wide range of potential users for this re-search project. Each of them has a handicap, which could be physical, cognitiveor sensorial. Also, as the user’s limitations have an impact on the HMI, thesystem must adapt to the user’s impairments and attempt to define how theHMIs should display notification messages in the most comprehensible way forthe user. Consequently, by taking these personas into account, we have defineda user profile (Figure 4) useful for the design of the system’s intelligence and theHMI. It is also flexible enough to deal with any types of users and will influencethe decision process.

Fig. 4. User profile

8

This user profile aims to consider cognitive and sensorial disabilities of theperson.

1. Inside user level we set four different grades: not possible (0) indicates thatthe user is not able to use the system - of course, this could be a temporarysituation. Easy (1), standard (2) and expert (3), indicates the user’s level ofunderstanding of the system. This understanding includes different issues:e.g. knowing all the system’s possibilities and features, having memory losses,technological skills, etc.

2. Interface indicates how the user will interact with the system. We distin-guish input (human’machine) and output (machinehuman). As input wehave voice control (most natural way of communicating helpful for thosepeople with reduced cognitive capacities or low technological skills), hapticcontrols (touch screens, keyboards, remote controllers, etc) and sensoric in-terfaces (inertial, pupil tracking, biosensor based systems, etc.) Output fromthe machine will normally be graphics, text and sound. Besides the user pref-erences, this could also have implicit information about the user’s cognitivelevel; if properly designed, graphical interfaces could be very intuitive.

3. Inside the audio category, important features are volume and pitch, keyissues that can help people with visual and aural disabilities to perceive theHMI.

4. Display includes common adjustment controls in many screens: contrast,brightness and colour settings. These characteristics, besides adapting tothe ambient light and user preferences, together with magnification, mighthelp people with visual impairments to interact with the display.

2.4 User profile for the Personas

Translation of the characteristics of the personas into the values in the user pro-file has been done considering the features and capabilities the persona relevantto the system. If we consider again the case of Hannah, the user profile could beas follows:

Hannah, 67 years old, SwedenUser level: Expert Interface: Input (haptic)

Output(graphics and sounds)

Audio: Volume (0-1) Display: Contrast (high)Pitch (normal) Magnification (high)

Brightness (high)Colour (2)

We were able to apply this user model to all ten personas, which makes us be-lieve that the model proposed can now be use in the architecture of the systembeing developed.

9

3 Use case: EasyLine + project

Elderly people suffer some disabilities that get worst with years. These disabil-ities make it more difficult to perform tasks in a normal independent life. It isa fact that the main disabilities prevent them carrying out domestic tasks andthat about a quarter of all household accidents occur in the kitchen where theappliances are key elements.

The principal objective of the EasyLine+ project is to develop an AmI kitchenwith advanced white goods prototypes near to market. This would increase theautonomy of the elderly and people with disabilities in their everyday activities,allowing them to live an independent life for a longer period of time. The sys-tem, being aware of the context and the user, enhances the intelligence of thewhite goods. With or without user cooperation, it will facilitate the use of theappliances, adapting the systems to the disabilities of the users. The system isalso a learning system meaning it can detect the user’s behavioural patterns andidentify any unusual changes or loss of abilities and try to compensate for them.Figure 5 shows a block diagram of the system.

Fig. 5. Block diagram of EasyLine+ system

4 Conclusions

In this system, user-awareness is of key importance. It has to know the usercapacities in order to help him/her adequately by adapting both cognitive and

10

HMI levels. For example, if the cognitive level is below the average, the interfacewill provide little information simultaneously, may use voice and graphics, elim-inate advanced options, etc. If the person is totally blind the HMI will use voiceor Braille communication. Context awareness is also necessary. It includes datafrom sensors (cupboards’ doors, presence, etc.), status of appliances and RFIDinformation from food in the kitchen, clothes in the washing machine, etc. Withall this information, the intelligent system will determine the commands sent tothe white goods and establish a bidirectional communication with the HMI.

The statistic predictions in Europe indicate a real global ageing of the popu-lation. This phenomenon will undoubtedly affect the market tendencies and willhave some repercussions in the information society. Thanks to the appearanceof many affine technologies in the home environment, it is easier to take stepstowards the creation of more friendly and useful domotic systems, which willconsequently increase the quality of life. Using a combination of the newest AmItechnologies together in the creation of user-centred design systems will be thekey for many applications oriented to a home environment.

This paper has argued the importance of user-modelling involvement in thedevelopment of an AmI system. Further testing on the results will attempt todemonstrate if the solution offers an opportunity for more vulnerable people(such as elderly people or persons with disabilities) to live independently athome for a longer period of time. The implemented system will monitor theenvironment and user’s behaviour by analysing the data gathered through thesensor network and the user inputs in a non-intrusive way; subsequently thesystem will make decisions so as to eventually send notifications (some requiringan action to be taken) and, if necessary, update the user’s profile to enhance theinteraction.

Nevertheless, there are some important issues, not discussed fully in this pa-per but necessary to mention. There are some significant risks of social exclusionor other ethical concerns related to smart environments. Therefore, AmI systemsin housing must support the user’s socialization, provide security and stimulateuser’s physical and mental activities [24]. As a result, innovative products needto satisfy certain requirements (e.g. affordability, user-friendliness, standards,robustness, interoperability etc.) in order to be accepted by a larger audience[25]. Another significant issue is the difficulties that software-based systems havein helping users with mobility impairments. Further improvements to solve thismay include the deployment of actuators, commuters and voice synthesis control.

5 Acknowledgements

This work has been supported by the EU FP6 ICT funded research project (no.045515), ”Low Cost Advanced White Goods for a Longer Independent Life forElderly People”.

11

References

1. Roy Delgado, A., Robinet, A., McGinn, J., Grout, V., Picking, R.: Human-MachineInterfaces for Smart Homes.In: 3rd Collaborative Research Symposium on Security,E-Learning, Internet and Networking (SEIN), pp. 147–157. Information Securityand Network Research Group, Plymouth (2007)

2. Cook, D. J.: Providing for Older Adults Using Smart Environment Technologies.IEEE USA Today’s Engineer Online 5(7), (2007)

3. IST Advisory Group: Scenarios for Ambient Intelligence in 2010. Final Report,EC. Brussels (2001).

4. Emiliani, P.L., Stephanidis, C.: Universal Access to ambient Intelligence Environ-ments: Opportunities and Challenges for People with Disabilities. IBM SystemsJournal 44 (3), pp. 605–620 (2005)

5. Belami Project, http://www.belami-project.org/definition6. Jorge, J.A.: Adaptive Tools for the Elderly: New Devices to Cope with Age-Induced

Cognitive Disabilities. In: 2001 EC/NFS Workshop on Universal Accessibility ofUbiquitous Computing: Providing for the Elderly, pp. 66-70. ACM 2001, Alcacerdo Sal (2001)

7. Gonalves, D.: Ubiquitous Computing and AI towards an Inclusive Society. In: 2001EC/NFS Workshop on Universal Accessibility of Ubiquitous Computing: Providingfor the Elderly, pp. 37-40. ACM 2001, Alcacer do Sal (2001)

8. Abascal, J.: Ambient Intelligence for people with disabilities and elderly people. In:SIGCHI Workshop on Ambient Intelligence for Scientific Discovery (AISD). ACM2004, Vienna (2004)

9. Cooper, A.: The Inmates are Running the Asylum: Why High Tech Products Driveus Crazy and How to Restore the Sanity. Sams Publishing, USA (2006)

10. Saffer, D.: Designing for Interaction: Creating Smart Applications and Clever De-vices. New Riders, Berkeley (2007)

11. Pruit, J., Adlin, T.: The Persona Lifecycle: Keeping People in Mind ThroughoutProduct Design. Morgan Kaufmann, San Francisco (2006)

12. Norman, D.: Ad-Hoc Personas and Empathetic Focus (2004)13. Eurostat, http://epp.eurostat.ec.europa.eu/14. EIAA, http://www.eiaa.net/15. Sleeman, D., Brown, J.S.: Intelligent Tutoring Systems, Academic Press, New York

(1982)16. Petrelli, D., De Angeli, A., Convertino, G.: A User Centered Approach to User

Modeling. In: 7th International Conference on User Modeling (UM99), pp. 255–264. Springer Wien, New York (1999)

17. Fink, J., Kobsa, A., Nill, A.: Adaptable and Adaptive Information Access for AllUsers: Including the Disabled and the Elderly. In: 6th International Conference onUser Modeling (UM97), pp. 171-173. Springer Wien, New York (1997)

18. Carroll, J.M., Rosson, M.B.: The Paradox of the Active User. In: Carroll, J.M.(eds.) Interfacing Thought: Cognitive Aspects of Human-Computer Interaction.pp. 80–111. MIT Press/Bradford Books, Cambridge (1987)

19. Barnard, P.: Interacting Cognitive Subsystems: A Psycholinguistic Approach toShort Term Memory. In: Ellis, A. (eds.) Progress in the Psychology of Language.pp. 197–258. Lawrence Erlbaum Associates, Hove (1985)

20. Johnson, C.W.: Decision Theory and Safety-Critical Interfaces. In: Nordby, K.,Helmersen, P.H., Gilmore, D., Arensen, S. (eds.) Interact ’95. pp. 127–132. Chap-man Hall, London (1995)

12

21. Kieras, D., Wood, S.D., Meyer, D.E.: Predictive Engineering Models based on theEPIC Architecture for Multimodal High Performance Human-Computer Interac-tion Tasks. ACM Transactions on Human-Computer Interaction (TOCHI) 4(3),230–275 (1997)

22. Shneiderman, B.: Universal Usability: Pushing Human-Computer Interaction Re-search to Empower Every Citizen. Communications of the ACM 43, 84–91 (2000)

23. Kules, B.: User Modeling for Adaptive and Adaptable Software Systems. In: ACMConference on Universal Usability. Arlington (2000)

24. Op Den Akker, R.: How Useful is an Intelligent Computer? In: 2004 Conferenceon Human Factors in Computing Systems (CHI 2004). ACM 2004, Vienna (2004)

25. Friedewald, M., Da Costa, O., Alahuhta, P., Heinonen, S.: Perspectives of AmbientIntelligence in the Home Environment. Telematics and Informatics 22(3), 221–238(2005)