Computers and Senior Citizens: Menu

God helps those who help themselves.

4. PROPOSED RESEARCH AND

PRELIMINARY WORK

This research focuses on elderly users that have a reasonable

proficiency in the use of computers. Our aim is to implement an

intelligent help facility to dynamically offer assistance when it is

recognised that a user may be experiencing difficulties. To do

this, the authors have carried out some preliminary work to

demonstrate that monitoring browser interaction patterns can be

effectively achieved and that Artificial Neural Networks (ANNs)

can be utilised to identify interaction patterns for individual users.

4.1 Using Intelligent Techniques

Initial experiments by the authors have established the usefulness

of an ANN to categorise users’ browsing tasks [14]. Data was

gathered from 20 volunteers. Each participant was asked to

complete two web browsing tasks, Task A and Task B. Each task

consisted of five small subtasks where the subject was asked to

find a specific piece of information by using the links on the

webpage specified. The aim was to identify which task a user

completed. An event logger (free software created by the Mozilla

project) was utilised to log participants’ activity as each task was

conducted [18] including the title and URL of the current

webpage, the time taken to complete the task and whether the left

or right key had been pressed on the mouse. A log was created in

the form of an XML document. A parser program was used to

extract the relevant data from the XML file. The information from

each logged event was stored in an Access database where the

data was analysed. After the data was gathered and analysed, half

of the data (seen data) was used to train the ANN. The experiment

was repeated ten times and after each repetition the classification

accuracy (CA) of the output was examined to calculate the

performance of the network. This was done by analysing the

number of correct classifications over the two tasks. Once the

accuracy of each training procedure had been established the

average was calculated and the overall CA determined. This was

repeated for different network architectures and once the optimum

network architecture (number of neurons in the hidden layer) had

been determined, the accuracy of the trained network was tested

by applying the remaining half of the data (unseen data) to the

network. Again, the process was repeated ten times, the average

obtained and the overall CA calculated. The overall CA obtained

was 96% when the ANN was tested using unseen data. This

preliminary work shows that ANNs can be successfully utilised to

monitor users’ browsing interaction patterns and thus identify

which task the user completed.

4.2 Monitoring Interaction Patterns

The initial step towards developing an evolvable interface is the

accurate identification of users’ browsing interaction patterns.

Initial studies have been carried out with users aged between 54

and 86 to establish typical interaction patterns when using the

Workshop "Supporting Human Memory with Interactive Systems", HCI Conference, September 4th, 2007, Lancaster, UK

- 30 -

Page 3

Internet. Data was gathered from a controlled study group of

elderly computer users and care was taken to select elderly people

with similar computer experience and age. Before the data

gathering process, the elderly participants were asked to complete

a questionnaire about their computer experience and the study

group was chosen based on the results obtained. Results from the

questionnaire established that the most common sites used by the

age group are shopping, hobby and email sites which help them to

keep in contact with the outside world. Common tasks carried out

on these types of websites were identified. Each task consisted of

a number of subtasks which required participants to navigate

through the website specified and therefore enabled the

interaction patterns of older users carrying out these tasks to be

collected. These preliminary experiments logged factors such as

time taken and the navigation pathways. The University of the

Third Age (U3A) has facilitated access to a wide population of

older people who are familiar with computers but who still require

assistance when accessing the Internet. The data gathered from

the study group will be interpreted to offer a level of assistance

that will suit individual needs. In addition, the elderly participants

were asked about any difficulties they encountered while using

the Internet. The most prominent answer was that they had

difficulty remembering how they had completed certain tasks,

such as opening emails, on previous occasions. The elderly

participants found that they could remember how to complete a

task when they were given individual assistance with small

prompts to jog their memory and help them to complete the task.