PERSONALIZATION IN INTERACTIVE SYSTEMS

The Generation of Compliance?

Stephen Sobol

Institute of Communications Studies, University of Leeds, UK

Keywords: Interactive Systems, Personalization.

Abstract: This paper examines applications of personalization in interactive systems and seeks to map success in this

area into a general theory of database and narrative. It is argued that conventions of human communication

established well before the digital age play a large part in determining user responses to personalization. The

analysis offers a logic to help determine when to apply personalization. The results of an experiment to

detect personalization effects are reported which provide evidence of the value of personalization.

1 INTRODUCTION

Today's interactive systems offer much scope for

personalization and a range of personalization

regimes have been implemented in different

circumstances, with different technologies and with

different objectives. Broadly, objectives of

personalization strategies typically include improved

ease of use and/or market effectiveness - see, for

example, Karat, Brodie, Karat, Vergo & Alpert,

2003.

Customization approaches, in which users are

able to alter the delivery of material, and adaptive

techniques which 'learn' from user behaviour and

adapt deliveries accordingly on the basis of some

model of user behaviour both provide for a richer

communication experience than can be achieved

with conventional, old, media. It may be, however,

that this richer experience is not always wanted:

"25% of consumers actually avoid personalized

websites because they fear that their personal

information will be abused." (McGovern, 2003).

Whilst there is an extensive literature on how to

personalize, customize and integrate adaptiveness

into interactive systems (Harris, 2002) there is often

the presumption that it is a good thing and if some

observe that it can be expensive and difficult

(McGovern, 2003) there is relatively little discussion

of the overall logic for personalization or analysis of

the situations in which it has been found to be

effective, ineffective or, indeed, damaging.

Since Manovich (2001) published 'The Language

of New Media' there has been much elaboration in

the literature on the relationship between database

and narrative in new media deliveries - see, for

example, Broden, Gallagher and Woytek 2004. We

also see the significance of narrative in software

development being raised by the advocates of 'agile

computing' - see Cohn, 2004.

The two-way flow of information enjoyed by

interactive systems permits a direct linkage between

databases and narratives and we can follow technical

debates about the mechanical and architectural

possibilities within such linkages (see, for example,

Instone, 2004). There is little doubt that user

interfaces can be improved enormously through a

deep appreciation of user needs (narrative) in system

development. This paper, however, seeks to argue

that personalization in interactive systems represents

the anticipation of narrative and however well it is

done, and however sensitive it is to collected data

there will be times when it is unwelcome to users -

and times when users will find it essential.

Reported here are the results of an experiment in

which regular users of an online forum exhibit

significantly different behaviours when faced with

personalized and non-personalized versions of a

questionnaire. Users, in a familiar situation, were

more likely to show 'compliant' responses when

faced with a personalized version of a questionnaire.

Effective personalization - in terms of enhancing

the user experience - may depend on the

circumstances. Established communication

139

Sobol S. (2006).

PERSONALIZATION IN INTERACTIVE SYSTEMS - The Generation of Compliance?.

In Proceedings of the Eighth International Conference on Enterprise Information Systems - HCI, pages 139-142

DOI: 10.5220/0002497301390142

 SciTePress

conventions outside the digital world may account

for variations in effectiveness.

2 COMMUNICATION

CONVENTIONS

In the context of the web we talk about users - not

viewers, listeners or readers. We might talk about

shoppers, clients, students, customers or

collaborators. A web user may be any of these things

at different times and at the technology level

communication is always, essentially two-way -

unlike television, radio or print media where

reception is typically anonymous. In these latter

media feedback is relatively loose, voluntary,

limited in specificity, reliant on respondent honesty

and subject to sampling error. According to Sobol

(Sobol and Stones, 2002) "The web offers us more

possibilities for data on real consumption patterns

than any other medium before it…". There is much

we can learn from careful analysis of these data

mountains - but because we can do things does not

mean that we should.

The communications technology that underlies

the web might be necessarily two-way - but that is a

matter for the devices involved and the protocols

that marshal the data that flows between them.

Human communication, on the other hand, takes

place via narrative. Whole industries and professions

have grown up which explicitly seek to convert data

to narrative (market research, medical research,

forensic science…) and back again (market strategy,

medicine, the law…). Because, in the case of the

web, the technical processes behind each conversion

are so closely allied we run the risk of ignoring the

established conventions of communication adopted

by, for example, viewers (passive receivers) and, for

example, customers (active participants in a

transaction). The thesis is then that personalization

strategies must take into account the communication

role of the user.

Communication conventions evolve. Security

cameras were once widely seen as a gross intrusion

but have become an accepted part of modern living.

If I have my newspaper delivered I know that I can't

expect to keep my address secret from the

newsagent - but if the vendor advises me to change

my paper I am outraged. The technology involved

in the transaction demands some participation from

me - some information from me - but only enough to

enable the sale. This is a data exchange. If I engage a

lawyer, or a doctor, my participation levels - and

information supply - will, typically, be much higher

and I will expect to receive advice and guidance

tailored to my personal circumstances. These are

narrative exchanges - and by the amount of

information I give I 'authorise' a personal response.

Collaborative working on interactive systems has

been found to be successful in circumstances where

levels of 'shared goals', 'trust' and 'impact' are high

(Stack, 1999; Sobol and Roux, 2004). It may be that

personalization can achieve some of its objectives by

accelerating development along these dimensions

but also the level of personalization that users find

acceptable may well be restricted by user scores on

the same factors. Personalization will 'work' if the

user submits, in principle, to guidance and

recognises its authority. In these cases mechanical

personalization may be able to steer a behaviour.

3 EXPERIMENTAL DESIGN

To explore these ideas users of a web-based

discussion forum at the Institute of Communications

Studies at Leeds University were given an online

questionnaire to complete on entry to the system

(The Forum). The system is used by staff and

students.

Access to The Forum is by username and

password - thus the system has the wherewithal to

deliver basic personalization. Users are welcomed

by name. Access to most discussion areas is

restricted and users only see areas (and can only

participate in areas) to which they have been granted

access. The system features a choice of methods

when it comes to entering a message. Users can

either have a plain HTML text box or can elect to

have a 'fancy text box' instead. The fancy text box

allows users to alter font sizes and colours as well as

apply bold and italic formatting. The resultant

HTML code can sometimes cause problems when

viewed in The Forum's talk areas. The system keeps

track of a user's preference in this regard and

delivers according to their last selection - always

allowing them to change.

On arrival members see a screen which shows

the most recent contributions to the areas open to

the user. Because of the large number of talk areas

on the system this can sometimes make the opening

page slow to load.

So, the system includes a very basic level of

personalization which does not prejudge - guess, or

deduce user requirements - rather it 'remembers' and

uses personal information (full name, email address)

openly to facilitate communication. The volume of

ICEIS 2006 - HUMAN-COMPUTER INTERACTION

140

traffic would indicate that users accepted the

convenience benefit consequent on the supply of

some personal information. A degree of trust has

been built up. The talk areas often carry PowerPoint

presentations used in lectures - providing 'impact' -

and the talk areas foster collaborative work of

various kinds which encourages the 'shared goals'

perception. For these reasons it is argued that the

system, however basic, has engaged users, to some

degree, in a relationship consistent with established

communication conventions in which the giving of

personal information is done for a reason and the

expectation of some return. If this is correct, and a

level of trust has been built up, we might expect that

trust to give some persuasive power to the system. If

personalization can be used as a way of exercising

that power and going beyond 'improved ease of use'

to 'improved market effectiveness' then it ought to be

possible to exercise that power in The Forum.

An online questionnaire was presented to all

Forum users on and between Monday 14th February

and Friday 18th February 2005. At random half of

the users received a personalized version of the

questionnaire - mentioning them by first name a

total of five times. The other users completed

questionnaires with no personalization. Users had to

complete the questionnaire in order to proceed to

'The Forum' proper. An example of a personalized

version appears below:

Hi there Steve, Forgive the intrusion but we would

like your opinion. You once tried the 'fancy text box'

and went back to the ordinary one. We'd like to

remove the 'fancy text box' because it can cause

problems with the web layout. What do you think?

[ ] Steve says leave it there as an option.

[ ] Steve says get rid of it.

Also, THE FORUM is slow to load. Mainly this is

because of the LATEST feature on the first page

which shows recent contributions in all areas.

Do you think its worth the wait - or should we get

rid of that feature?

[ ] Steve says leave it as it is.

[ ] Steve says get rid of the LATEST feature.

How would you feel about this system holding your

examination marks so that only you could see them

after logging in to the intranet?

[ ] Good idea. Very convenient - or other reason.

[ ] Bad idea. I'd be worried about someone hacking

in to my private area - or other reason.

Do you think THE FORUM should be opened up to

former students?

[ ] Yes. [ ] No.

The first question - involving the phrase 'we'd

like to…' is designed to test whether the

personalization could be associated with compliant

behaviour. Would those answering personal

questionnaires be more likely to comply with the

'system's' desire to 'get rid of' the fancy text box?

The second question is not directly loaded - we

might expect personalization to be less significant

here. Remaining questions were not personalized in

the expectation that any personalization effect would

be reduced. The third question was designed to give

a concrete indication of the level of trust that had

been built up among users and answer a local need.

The fourth question relating to former students

was, again, designed to answer a local need - but any

reluctance to allow 'strangers' into the trusted zone

might be seen as a measure of the strength of the

relationship between the system and the user.

4 RESULTS

A total of 234 responses were achieved over the five

days. A crosstabulation of the responses to question

1 against the personalization variable is given below

(expected values in parentheses) in Table 1. Those

who got the 'personal approach' were more likely to

exhibit compliant responses - i.e. to prefer the

removal of the 'fancy text box'.

Table 1: Personalization and preferences for the 'fancy text

box'.

Stay as is Get Rid

Impersonal Q

71 (63) 49 (57) 120

Personal Q

51 (59) 63 (55) 114

122 112 234

A chi-square analysis of this data applying

Yates's correction reveals the table to be significant

at 5% and the larger number of personalised

questionnaires yielding 'get rid' answers than one

would expect can be taken as evidence of a

personalization effect generating compliance. The

63 users who received the personalized

questionnaire and also voted to lose the 'fancy text

box' is larger than the 55 that would have been

expected by chance to the extent that there is only a

5% probability that the observed figures are the

result of chance. Personalization in the question is

significantly associated with compliance in the

answer. Table 2 provides evidence of a

personalization effect in respect of views on whether

or not to admit alumni to the discussion area. Those

PERSONALIZATION IN INTERACTIVE SYSTEMS - The Generation of Compliance?

141

who received a personalized questionnaire were

more likely to prefer the exclusion of alumni.

Table 2: Personalization and preference on the alumni

question.

Admit

Alumni

Do not admit

alumni

Impersonal Q

93 (86) 27 (34) 120

Personal Q

74 (81) 40 (33) 114

167 67 234

The table is significant at 5% - the personal

approach is associated with the preference of barring

alumni from the discussion areas. It appears that the

personalization here has engendered a feeling for

privacy in the relationship between the user and the

system. The general notion of the importance of

established communication conventions, involving

the build up of trust, in web systems has been

supported by the data. Further, we find that trust can

be exploited through personalization.

It should be remembered that 'The Forum' is part

of the day-to-day lives of the respondents. It

provides information and discussion of immediate

value (impact) to users and as such has built up a

certain level of trust (those 'publishing' on the

system rely on it to deliver) among people with

shared goals. It may be that the personalization

effects observed here are particular to such

situations. In this sense 'earned' personalization may

offer potential rewards whereas the unearned variety

may have more risks associated with it. All of which

would be consistent with the communication

conventions outlined in section 2 above.

5 CONCLUSION

Web systems need to respect established

communication conventions. Therefore as these

change so should the systems - the web itself may

change some conventions. In the same way that

many of us now accept the principle of closed circuit

television (in return for the security benefits) the

convenience afforded by web-site personalization

may become accepted, tolerated and even desired.

And then assumed. Personalized mailings may once

have won elections but are now barely noticed. It

may be that interactive systems are currently

enjoying a persuasiveness dividend by virtue of their

youth, but the fundamental point relating to the give

and take involved in the development of trust in a

communication relationship remains. Personal

information used in web systems is likely to be most

effective if that information would be used in a truly

personal interaction.

In summary, this paper provides evidence of a

personalization effect consistent with established

protocols of human communication in which

personalization reinforces an existing strong

communication relationship.

REFERENCES

Broden, N., Gallagher, M. and Woytek, J. (2004) Use of

Narrative in Interactive Design. Retrieved 21 January

2005 from

http://www.boxesandarrows.com/archives/use_of_nar

rative_in_interactive_design.php

Cohn, M. (2004) User Stories Applied: For Agile Software

Development. Addison-Wesley

Instone, K. (2004) An Information Architecture

Perspective on Personalization in Karat, C.M., Blom,

J.O., Karat, J. (Eds.) Designing Personalized User

Experiences in eCommerce. Springer.

Harris, N. (2002) Personalisation Literature Review.

Authenticated Networked Guided Environment for

Learning. Retrieved from

http://www.angel.ac.uk/public-

files/pdf/personalisation.pdf 19 February 2005

Karat, C.M., Brodie, C., Karat, J., Vergo, J. and Alpert,

S.R. (2003) Personalizing the user experience on

ibm.com. IBM Systems Journal 42 (4) 686-701

Manovich, L. (2001) The Language of New Media, MIT

Press

McGovern, G. (2003). Why Personalization hasn't worked.

Retrieved 22 January 2005 from

http://gerrymcgovern.com/nt/2003/nt_2003_10_20_pe

rsonalization.htm

Sobol, S. and Roux, J. (2004) Developing Web-Based

Infrastructures For Collaborative Working: The Case

For Localised Solutions IEEE Database and Expert

Systems Applications Conference

Sobol, S. and Stones, C. (2002) DMASC: A Tool for

Visualizing User Paths Through a Web Site Third

International Workshop on Management of

Information on the Web in conjunction with 13th

International Conference on Database and Expert

Systems Applications DEXA'2002, 389-393

Stack, J.M. (1999) Collaboration: Success for the Future.

InForum’99 - Proceedings of the DOE Technical

Information Meeting.

ICEIS 2006 - HUMAN-COMPUTER INTERACTION

142