Existing Zara Response Data 

Common Derailing/Difficult Topics

Relevant

(Functional but out of scope)

Irrelevant

(Socially inappropriate, existential/predictive questions)

Negative

Neutral

Positive

Strategic mitigation responses 

1 condition with 4 factors (baseline inclusive)

or

2 * 2 Matrix Study 

(modeled after "Gracefully mitigating Breakdowns in Robotic Services")

*Need good reasons to justify the factors we control

Control groups: Robotic response and no challenging questions posed

What is the purpose or areas of application for Zara? 

-- Caretaker paradigm or assistant/companion paradigm

 (From "Socially Intelligent robots: dimensions of HRI")

-- Important to direct the formation of strategies 

Will Zara have a physical presence in the future?

-- Virtual vs Physical 

-- Important to experimental design of whether to include another factor with 2 conditions

(humanlike vs virtual robot) 

What is the purpose or areas of application for Zara? 

-- Now: Personality assessor 

-- Future:  Ambassador/Receptionist role at UST

Will Zara have a physical presence in the future?

-- For the extent of experiment: a virtual entity

-- In the future: Physical entity

How can Zara detect the relevant or irrelevant questions?  

-- Relevant: "factual questions" or "keyword" 

-- Irrelevant: Socially inappropriate (Sexually lewd, swear-words like "f*** you", "bitch", "whore" that are commonly associated with women), 

Existential questions (keywords "life", "death", spiritual, religion, values) 

Paper Hunt on Current Way and Challenging Topics people impose to break/challenge robots

-- Assumption: Intentionality. What about times when people don't understand "limits" of robot? 

"Gracefully"

-- Subjective interpretation

-- Understand the psychology and qualities of what people perceive as graceful 

If data available, organize and find patterns of most common question type or topic

184 responses in total

65 >= State1

16 challenges

Oct - Dec 2015

Seeking reciprocity from Zara

16 unique ID responses, 124 responses in total to all 15Q Zara posed. Of the 30 challenging responses, ...

8

Stated verbal avoidance of topic

5

Implied avoidance of topic

8

2

Abusive Language

4

Seeking Clarification

3

Deliberate Challenge of Zara's ability

13

Seeking reciprocity from Zara

8

13

Avoidance of topic

Bonding

Mutual trust

Psychological need of Users

Intuition: Humans needs cues to bond with robots the same way humans bond with each other 

Context: Most of the responses are from Asian users

How reliable are these insights?

-- 16 challenges/180 responses is a small sample size

-- Lee, M.K. et. al. "Gracefully Mitigating Breakdowns in Robotic Services" (2010) : 457 respondents to survey

How might Zara respond to challenging questions/derailing topics in an empathetically graceful manner that enhances mutual trust and information reciprocity?

-- Based on the psychological need of the users

-- Ease into conversations

Discussion with Prof Ma 

-- Do we need more response data? 

-- If yes, what setting and what factors should we control?

-- Intuition: Need more data

-- Gather more data from students overseas (international pool to balance out the Asian users)

Mutual Trust Assumption Validation 

-- Psychology literature on mutual trust and bonding

-- A. Aron et. al. "The Experimental Generation of Interpersonal Closeness"  (1997)  

Paper Submission Deadline: 18 Sep 2016

Conference Date: 6-11 May 2017

Location: Denver, USA

*According to: http://www.sigchi.org/conferences

A. Trust: "the attitude that an agent will help achieve an individual’s goals in a situation characterized by uncertainty and vulnerability"

B. Cognitive Model: Human trust in an agent involves an attempt to construct an ad-hoc theoretical model about agent behavior 

C. Human trust in a non-human agent may also similarly be dependent on the degree to which the agent is perceived to possess such intentionality. 

D. Key results: 1. Participants showed more calibrate trust with immediate response from robots; 2. malevolent human-like agents are trusted less compared to machine-like agents; 3. machine-like agents are trusted more than human-like agents because of an automation bias

Objective: Evaluate and quantify the effects of human, robot, and environmental factors on perceived trust in HRI based on 29 empirical literature (1996-2010) with experimental and correlational data. 

Key findings: Robotic performance and attributes > human factors > environmental factors 

Failure Example: SWORD (Human + Robot ==> Failed to accomplish same goal)

Importance to Zara: 1. Higher trust is correlated with higher reliability (Ross, 2008)

2. Robotic Behaviorial Design 

Need more research; analysis tapped the surface in the paper. 

Avoidance is a coping mechanism in response to some form of stressor, such as fear, discomfort, or anxiety 

Suggests attempt to protect oneself from experiencing psychological damage

Insight: Two types of coping mechanisms active approaching (explicitly stating discomfort/doing sth to reduce stress) and passive avoidance (neglect, suppression, silence). This validates the categories for passive and active avoidance of topics

Implications: Instead of active and passive avoidance, change active to "Approach" and leave passive as "Avoidance" 

Approach-Avoidance Dimension by Shrontz (1975)

10 Total

2 Reported Aborted Attempts

4 Successful Attempts

What did you like about Zara? 

What did you not like about Zara? 

Where can Zara improve? 

- One of its kind

- Interesting

- Piqued curiosity

- Felt Spooky (chrome warning < microphone < cam)

- Off-beat stares and pose of avatar

- Monotone replies 

- Q are too personal

-Chinese Version: "did not understand"

- Informational Landing page of some sort (prepare the users) 

Terse AND Verbose responses: 

"N" scores is way off the charts (3000 pts)