Simmons, R., Makatchev, M., Kirby R,. Lee, M., Fanaswala, I., Browning, B., Forlizzi, J. & Sakr, M. (2011). Believable Robot Characters. AI Magazine, winter 2011, 39-52.

@Article{SimmonsMakatchevKirbyLeeFanaswalaBrowningForlizziSakr2011,
  author = 	 {Simmons, R., Makatchev, M., Kirby R,. Lee, M., Fanaswala, I., Browning, B., Forlizzi, J. & Sakr, M.},
  title = 	 {Believable Robot Charcters},
  journal = 	 {AI Magazine},
  year = 	 {2011},
  volume = 	 {winter},
  pages = 	 {39--52}
}
 

Author of the summary: Hana Lang, 2011, hanarae@gmail.com

Cite this paper for:

• evaluating robot characters in the wild [pg46]
• autonomous robotic systems capable of natural and engaging social interactions with untrained users [pg39]
• "roboceptionists" at Carnagie Mellon campuses in Pittsburgh, Pennsylvania and Doha, Qatar [case studies throughout article]
• use of look, voice, personality, language, nonverbal expressions, backstory, evolving storyline to achieve overall engaging experience as per the acting theories of Delsarte (Stebbins 1886) and Stanislavs (2008) [pg41]
• Drama school collaboration to create characters with depth and context [pg42]
• Backstory elements including age, family, living situation, past employment etc. to create believable robotic agents [pg 42]
• use of objects to represent life events in robot backstory [pg42]
• each object may have time line attribute, allowing a user to access event information by way of that events object and assigned attributes [pg42]
• attraction of users as being a main goal of robots in the wild [pg43]
• interaction predictability using position, range, and velocity of user in relation to robot [pg43]
• mood representation as valence and intensity in accordance with social exchanges (Rook, 2001) [pg44]
• categorical theory of emotion (Gockley, Forizzi, and Simmons 2006; Kirby, Forzilli and Simmons, 2010) applied to robot character [pg44]
• autonomic behaviour in artificial intelligence [pg44]
• cultural identity implementation in artificial intelligence using language, backstory, cultural cues [pg45]
• use of richpoints (important cues discovered through anthropological, ethnographical, second language acquisition investigations) to encourage homophile (Lazaesfeld and Merton 1954) between robot and user (Nass, Isbister, and Lee 2000) [pg45]
• presence of user greeting robot as being indicative of interaction length[pg44]
• users who greet robotic agents and use social dialogue are better at getting questions successfully parsed and answered [pg44]
• robots who use techniques such as priming more likely to be thanked/ be granted a farewell by users [pg46]
• robotic agents that exercise non verbal expressions of effort (NVEE) more likely to be thanked / be granted farewell [pg48]
• pragmatic features such as implicature computation contribute to robot believability [pg49]

href="http://find.galegroup.com.proxy.library.carleton.ca/gtx/infomark.do?docType=IAC&prodId=AONE&tabID=T002&searchType=PublicationSearchForm&type=retrieve&version=1.0&userGroupName=ocul_carleton&docId=A276517810&contentSet=IAC-Documents&source=gale&infoPage=infoMarkPage

What makes a robot character believable? Richness and continuity of character. How are these qualities achieved?

Backstory [pg42]

Creating evolving story line that includes details about character's life.

Events are designed and scheduled in advance.

The events contribute to characters's mood and emotion.

Events are created as objects in the character's internal content database.

Each event has attributes including a time for which it starts and ends, and associated mood, dialogue, and a list of any other of the character's internal objects that may be involved in the event.

Social Verbal and Nonverbal Behaviours: [pg43]

Socially appropriate behaviours may facilitate attracting interactors.

AI scientists trained a classifier to predict how likely a person will be to interact with the robot based on proximity, velocity etc. If the classifier determines a person is likely to interact, the robot will initiate the interaction.

There are also strategies used to encourage social responses.

Facial expression and gaze contribute to liveliness, which in turn promotes social responses such as thanking and issuing farewells.

Robot spontaneously engages in autonomic behaviour such as breathing.

Mood: [pg48]

Mood represented as a valence and intensity. Positive social exchanges improve mood of robot, negative exchanges cause mood to worsen.

Expressing Culture: [pg45]

Users more likely to interact with robots if they experience sense of sameness, known as homophily.

Using language, context and appearance, engineers create robots in image of the majority demographic of interactors.

Topics, Greeting, Discourse features: [pg46]

Two classes of users: those seeking information, and those inquiring about the robot itself.

Eliciting Social Behaviours through Dialogue Strategies: [pg47]

Greeting an interactor increases likelihood of user responding with a greeting.

Those who greet the robot more likely to get their questions answered. Users more likely to thank robot if the robot makes a non-verbal expression of effort.

Pragmatic Analysis: [pg49]

Important feature of natural language that needs to be addressed in AI are the implicatures that are generated in natural speech.

Figure 1 is a photo of Valerie at Work at Carnegie Mellon in Pittsburgh, Pennsylvania [pg 41]
Figure 2 is a photo of Tank at Work at Carnegie Mellon in Pittsburgh, Pennsylvania [pg 42]
Figure 3 is a photo of Hala at Work in Doha, Qatar [pg 43]
Figure 4 is a photo of the faces of all three roboceptionists, including the new version of Valerie [pg44]
Figure 5 depicts Valerie's emotional range [pg45]
Figure 6 depicts the Temporal Flow of an information-Seeking Interaction [pg47]
Figure 7 depicts a dialogue fragment containing the embedded question in Turns 2 and 4, to illustrate implicature construction[pg50]

Table 1 depicts frequency of interactions by topic adopted from Lee and Markatchev (2009)

Summary author's notes:

• the page numbers are from the online version of AI magazine.