Towards Caring Machines (Bickmore & Picard)

Bickmore, T. W. & Picard, R. (2004). Towards Caring Machines, CHI 2004, Late Breaking Results Paper, 24-29. Vienna, Austria.

This paper examines interface behaviors that elicit the perception of caring. Results from a longitudinal study of simulated caring by a computer are presented, in which 60 subjects interacted with a computer agent daily for a month, half with a “caring” agent and half with an agent that did not use behaviors to demonstrate caring. The perception of caring by subjects in the “caring” condition was significantly higher after four weeks, and was also reflected in qualitative interviews with them, and in a significantly higher reported willingness to continue working with the “caring” agent.

“In education, it is known that the presence of someone who is perceived as caring can be motivating [21], and various studies have also linked caring and other qualities of interpersonal relationships between teachers and students to motivational outcomes over the long term [5]” (p.1489).

There are several behaviors an interface agent or robot could use to elicit the perception of feeling cared for by a user: (see [3] for a summary)

  • demonstrations of empathy and comforting behavior
  • social dialogue
  • self disclosure
  • emphasizing commonalities
  • meta-relational communication (particularly emotional aspects)
  • talking about the past and future together
  • continuity behaviors (appropriate greetings and farewells and talk about the time spent apart)
  • reference to mutual knowledge
  • explicit messages of esteem
  • Nonverbal
    • facial expressiveness (including displays of concern)
    • head nodding
    • tone and timing of speech.
  • Nonverbal “immediacy” behaviors
    • close conversational distance
    • direct body and facial orientation
    • forward lean
    • increased and direct gaze
    • frequent gesturing and postural openness

Although there have not been any studies to date on the ability of computer agents to affect the feeling of being cared for, there have been many studies on the ability of computer agents to affect related psychosocial constructs, such as liking of and trust in the agent.

  • Reeves and Nass demonstrated the ability of computers to increase users’ liking of them through the use of flattery, praise of other computers, matching the user in personality, or the use of “in-group” cues [18].
  • Morkes, Kernal and Nass demonstrated that computer agents that use humor are rated as more likable, competent and cooperative than those that do not [15].
  • Embodied pedagogical agents, especially those that are highly expressive, have been found to increase students’ perceptions of trust: such agents are perceived as helpful, believable, and concerned [13].
  • Mulken, et al, found that personification of an interface by itself does not appear to be a sufficient condition for raising the trustworthiness of a computer [16]. In an experiment with REA—a life-sized, animated virtual real estate agent—Bickmore showed that the agent’s use of social dialogue increased trust in it for extroverts (for introverts it had no effect) [4]. [Likewise, some people responded positively to “relational-Laura” and some didn’t…it’s a complicated story.]

An agent named Laura played the role of an exercise advisor designed to help subjects through a behavior change program designed to increase their physical activity levels. The agent appeared as an embodied conversational agent. Subjects conducted a 10 minute interaction with Laura daily on their home computers for one month, during which Laura provided feedback on their exercise behavior, helped them overcome obstacles to exercise, provided educational content related to exercise, and obtained and followed up on commitments to exercise.

A RELATIONAL version of the agent used all of the caring behaviors described above. For example, if a subject indicated they were not feeling well (and thus unable to exercise), Laura provide appropriate empathetic feedback while exhibiting a concerned facial expression.

A NON-RELATIONAL version of the agent delivered identical health content but had all caring and relational behaviors removed.

The principal outcome measure used in the study was the Working Alliance Inventory, a 36-item self-report questionnaire used in psychotherapy that measures the trust and belief that the therapist and patient have in each other as team-members in achieving a desired outcome [11].

When asked at the end of the month if they would like to continue working with Laura, subjects in the
RELATIONAL condition also responded much more favorably than the NON-RELATIONAL group, t(57)=2.43, p=.009.

One behavioral measure related to caring was evaluated. In the closing session, subjects were given a choice of farewell greetings to say goodbye to the agent. Significantly more subjects in the RELATIONAL group (69%) chose the most sentimental farewell (“Take care Laura, I’ll miss you.” vs. “Bye.”) than in the NON-RELATIONAL condition (35%), t(54)=2.80, p=.004.

Post-study feedback and reflection on the experience “illustrate a range of user feedback about a system that might evoke feelings of caring – from liking to disliking, from acceptance of the effects to denial of any effects” (p.1491).


3. Bickmore, T. Relational Agents: Effecting Change through Human-Computer Relationships. PhD Thesis, Media Arts & Sciences, MIT, Cambridge, MA, 2003.

4. Bickmore, T. and Cassell, J., Relational Agents: A Model and Implementation of Building User Trust.
Proc. CHI 2001, CHI Letters 3(1), 396-403.

5. Birch, S. H. and Ladd, G. W., Interpersonal Relationships in the School Environment and Children’s
Early School Adjustment. In J. Juvonen and K. Wentzel, Eds., Social Motivation: Understanding Children’s School Adjustment, Cambridge University Press, NY, 1996.

11. Horvath, A. and Greenberg, L. Development and Validation of the Working Alliance Inventory. Journal
of Counseling Psychology, 36, 2 (1989) 223-233.

13. Lester, J. C., Converse, S. A., Kahler, S. E., Barlow, S.T., Stone, B. A., and Bhogal, R. S. The Persona Effect: Affective Impact of Animated Pedagogical Agents. Proc. CHI 1997, 359-366.

15. Morkes, J., Kernal, H., and Nass, C. Humor in Task-Oriented Computer-Mediated Communication and
Human-Computer Interaction. Proc. CHI 1998 215-216.

16. Mulken, S. v., Andre, E., and Muller, J. An Empirical Study on the Trustworthiness of Life-Like Interface Agents. Proc. HCI 1999, 152-156.

18. Reeves, B. and Nass, C. The Media Equation. Cambridge University Press, Cambridge, 1996.

21. Wentzel, K. Student Motivation in Middle School: The Role of Perceived Pedagogical Caring. Journal of Educational Psychology, 89, 3 (1997) 411-419.

2 Responses to “Towards Caring Machines (Bickmore & Picard)”
Check out what others are saying...
  1. […] Here is the original: Towards Caring Machines (Bickmore & Picard) […]

  2. […] This post was mentioned on Twitter by Chris Brogan, missdestructo. missdestructo said: RT @chrisbrogan: Someone is working on replicants perhaps? – […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: