04.11.2014 Views

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

elektronická verzia publikácie - FIIT STU - Slovenská technická ...

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

266 Selected Studies on Software and Information Systems<br />

tunities for self-expression and self-disclosure, physical support, and reassurance of worth<br />

and value. Different models of relationship were proposed.<br />

In social exchange models, costs, benefits and investments in "relationship business" are<br />

modeled (Burke, 2006), i.e. the perceived costs and benefits of providing and receiving<br />

provisions respectively, and whether suitable alternatives are available affect the relationship's<br />

duration. In dimensional models, features that characterize different stereotypical interactions<br />

are used. (Svennevig, 1999) proposes an extension of earlier models into four dimensions<br />

of relationship as follows:<br />

1. Power. The person's ability to control the behavior of the other. Typically, occurs<br />

when the distribution of rights and obligations resulting from the relationship roles is<br />

asymmetric, e.g. institutional roles and positions.<br />

2. Solidarity. The degree of like-mindedness, e.g. political membership, religion, gender,<br />

birthplace. Rights and obligations are usually symmetrically distributed.<br />

3. Familiarity. Mutual knowledge of personal information, i.e. breadth (number of topics)<br />

and depth (public or private) of information disclosed.<br />

4. Affect. Mutual attraction, i.e. the degree of liking for each other.<br />

These dimensions are usually interrelated, though it needs not to be the case. For example,<br />

high levels of solidarity based on group membership need not to elicit high levels of either<br />

familiarity or affect, and pure affect arises in some situations (e.g. love at first sight).<br />

People engage in different relational strategies for maintaining relationships. Five<br />

strategic maintenance behaviors that are used intentionally to keep up the relationship were<br />

identified (Haas, 2005): positivity (e.g. cheerfulness, positive comments), openness (e.g. selfdisclosure,<br />

meta-relational communication), assurances (e.g. verbal or nonverbal expressions<br />

of love and comfort), shared tasks (e.g. household duties), and social networks (e.g.<br />

communicating with mutual friendships). In addition to these, different routine maintenance<br />

behaviors that are regular or habitual behaviors that serve to maintain the relationship<br />

were identified, such as joint activities, affection, avoidance of conflict, and focus on self<br />

(e.g. watching weight, furthering career). Finally, people expect relationships to change<br />

over time (Duck, 2007) i.e. the content, quality and diversity of interactions and activities<br />

change, reciprocal behavior decreases and complementary behavior increases, among<br />

others.<br />

With language being the principal means for developing relationships, relational<br />

agents need to implement the relational strategies outlined above in dialogues with their<br />

users. Pioneering work in this area is the computational model of mixed social-and-task<br />

dialogue addressing some of these strategies (Bickmore, 2003) evaluated within the REA<br />

system in which an embodied conversational agent (ECA) performs the role of a real-estate<br />

salesperson. Within this domain, the real-estate agent pursues several goals in parallel:<br />

determining clients' housing preferences (e.g. size, location), establishing trust and reducing<br />

clients' fear about such a big purchase, establishing agent's expertise and credibility.<br />

Although REA has a fully articulated graphical body, senses the user through cameras and<br />

microphones in real-time, and is capable of speech with intonation, facial display and gestural<br />

output, it is her conversational properties that are interesting to us. Indeed, later in<br />

evaluation it was found out that REA's nonverbal behavior was utterly insufficient (even<br />

inappropriate), and specifically her conversational properties exhibited in a second phone-

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!