Towards Extracting Personality Trait Data from Interaction Behaviour: Towards Extracting Personality Trait Data from Interaction Behaviour Nick Fine and Willem-Paul Brinkman
School of Information Systems, Computing and Mathematics
Keywords: logging, log file recording, user interface skins, reskinning, user interface design
Problem 1: Avoiding Average: Problem 1: Avoiding Average Average user interfaces = average interaction
Why interact with a UI that is designed for the
UI skinning technology allows for easy change
of the UI – but how can this best be achieved?
Problem 2: Segmenting Large User Populations: Problem 2: Segmenting Large User Populations If not designing for average, need to target
certain subsets of the larger population:
how are they identified?
how are they designed for?
Approach: Approach 1) Segment large user populations by a defining trait – Personality
why Personality? CASA (Reeves and Nass), Similarity Attaction Hypothesis (Byrne and Nelson), Colour Theory
2) Determine Personality through log file recording
Informs designers of the Personality types of the target population without needing to ask the users directly
3) Produce Profiled User Interface Skins (ProSkins) that are designed for the target segment
e.g. Red UI skin colour for extroverts
e.g. Low edge complexity for introverts
e.g. Agreeable personality represented for Agreeable users
Towards the Individual: Designing for Subsets: Towards the Individual: Designing for Subsets Log File Recording: Log File Recording Segment profiles established using log file recording
methods to capture:
User interactive behavioural measures
Mouse clicks (navigation, feature use, sessions)
UI Skin selection
User questionnaire data
Personality (IPIP-NEO, TIPI)
UI Skin Preference
Music Preference (STOMP)
General Demographics (age, gender, country)
Experimental Platform: Infrastructure: Experimental Platform: Infrastructure Client-Server over TCP/IP
Microsoft .NET 1.1 Framework
Access Database Experimental Platform: Application Architecture: Experimental Platform: Application Architecture Analysis: Analysis Looking for relationships between user
Personality and recorded interactive
'Big Five' (Costa and McRae)
Openness to new experience
(as measured by the IPIP-NEO)
Number of events in session (N, M, SD)
Total events of all sessions
Correlation – events and N sessions
Results: Results Ethical Issues: Ethical Issues Slide12: Position Statement: Position Statement In order to provide personalisation and customisation
services greater information about users is required.
Log File Recording (LFR) provides a means to collect this
information in an unobtrusive manner.
How can HCI develop LFR as a research method within an
Issues: Issues What kind of information is acceptable to record?
Personally/non personally identifiable?
e.g. Age, Gender, Country
e.g. WWW sites visited
e.g. User interface skin choices/configuration
e.g. Personality, Intelligence, Cognitive Style
e.g. application usage, feature usage, mouse clicks, number of sessions, mean times
Issues: Issues Is it acceptable/possible to record data
which can then be used to identify the
If the data recorded is not personally
identifiable, what potential harm is there?
If no harm, then why the need to disclose?! Logging is Already Ubiquitous!: Logging is Already Ubiquitous! Log file recording is and
has been recording
user interactive behaviour
web server logs
media player content
any IP access
door security systems
content management systems
license plate recognition systems
Protecting Users: Protecting Users Anonymity
Not personally identifiable, therefore no risk to individual privacy
Full disclosure and permission
Ability to view logged data and/or source code
The 'open source' philosophy
Ability for user to turn off logging/opt out
User has option to withdraw from logging at any time
Giving to Get: Giving to Get How can we gain trust and overcome user
scepticism regarding LFR?
If users perceived usability data derived
from LFR as harmless then more people
would contribute usability data freely.
SETI@home, folding@home, distributed.net