alina pommeranz, msc in interactive system engineering supervised by dr. ir. pascal wiggers and...
TRANSCRIPT
Alina Pommeranz, MSc in Interactive System Engineeringsupervised by Dr. ir. Pascal Wiggers and Prof. Dr. Catholijn M. Jonker
Content
Introduction Pocket Negotiator Research Goals from HCI perspective First Experiment User-centered Approach Questions and Suggestions
About myself
Education 4-years Computer Science, FH
Gelsenkirchen, Germany 2-years Master of Science in Interactive
Systems Engineering, KTH Stockholm, Sweden
Previous Research 1 year at the Swedish Institute of
Computer Science (SICS)
Negotiation support
Negotiation is a common task,but still only few people are good negotiators
Computers can better cope with the computational complexity (many options, combinations of variables, solutions)
Negotiation cannot be handled by artificial intelligence alone (semantic problem , emotional issues involved)
Pocket Negotiator(PN) “new type of human-machine
collaborative system that combines the strengths of both [human and system] and reduces the weaknesses.” (Jonker, 2007)
The PN will handle computational complexity issues provide bidding- and interaction advice
the user will handle background knowledge interaction with the opponent negotiator
Pocket Negotiator (PN)
handheld device 2 domains:
real estate job negotiations
for non-expert users
Issues from a Man-Machine Interaction point of view Preference Elicitation
User preference model Interaction between the system and
the user (generic task model) Explanation/Training Module (use of
animated character) Visualization techniques (for small
devices)
Preference Elicitation Experiment Goal: investigate preference elicitation
techniques involving different research areas (AI, HCI, Affective Computing)
Evaluation of user satisfaction (liking, ease of use, intuitivity) with different ways of giving preferences
Holiday domain 32 participants 8 ordering/rating tasks (including affective
feedback and navigational interaction), 2 comparisons and a questionnaire
Lexicographic ordering Paper submitted to HuCom 2008
Experimental setup
activeHotell Active
Results
Traditional property rating seems most preferred and resulted in one of the best orderings of the outcomes space when using the lexicographic algorithm
Properties of holidays are interdependent for more than a 3rd of the participants
Considering affective attitude can improve understanding of users’ preferences
With a multi-angle approach we could identify limitations and issues that would have not been found with just one research perspective
How to use the results?
When designing a new preference elicitation method we need to consider dependencies of
preferences navigation task is suited to find a number of
such dependencies rating of properties should be done in
known and preferred ways (Likert scale) we could incorporate ways to express affect
to properties/outcomes a combination of different techniques might
be the way to go (needs to be tested)
Limitations of the experiment Complete ordering of holidays as
control condition small set of properties and their values Rating and ordering tasks very quick and
with little effort Strengths of other techniques in the
background
User-centered approach I
Video-supported study on social acceptance, practices and use-contexts (scenarios)
Case studies with potential users including Observation of real life negotiation situations (not
only bidding but all other phases, too) Capturing such situations on camera Maybe physiological measures (heart rate, GSR
etc. to measure emotions) Questionnaires and interviews before and after
the negotiation phases Diaries by the users
Interviews and Focus groups with experts in the domains (e.g. real estate agents, P&O members)
User-centered approach – II Design of Negotiation Scenarios based
on the case study data and used in further experiments
Iterative implementation of prototypes Wizard-Of-OZ experiments to test
(with prototypes): Preference elicitation techniques User-system interaction techniques Visualization techniques Explanation/ training module
Questions? Suggestions?
Thanks for your attention!