Kinda Human

Dates: August 2018 to May 2019
Team: Scott Dombkowski
Advisors: Stacie Rohrbach and Molly Wright Steenson
Work Type: Academic

Kinda Human is a thesis project on enhancing expression and understanding in an intimate relationship through artificial agents.


Today, many designers concern themselves with efficiency, effectiveness, and productivity when considering relationships between humans and artificial agents. While these three dimensions are essential to consider, they fail to embrace our humanity (i.e., our ability for self-expression and for understanding others, the emotional and intellectual mechanisms we employ, the vast differences in our makeup and experiences, our propensity to make the same errors more than once, and our idiosyncrasies) and present an opportunity for designers to consider dimensions present in our relationships that make us human. This thesis project explores two of those dimensions—expression and understanding—in the context of intimate relationships.

In this thesis, I:

  • Investigated the theory, potential applications, and affordances of intimate relationships (which tend to be a person's most defining and determining), artificial agents, and other relevant research areas.
  • Acquired a comprehensive understanding of the models users have of artificial agents and intimate relationships by conducting several design studies.
  • Created and evaluated several artificial agents designed to enhance intimate partners' capacity for expression and understanding.
  • Developed a set of principles and challenges to assist those designing for expression and understanding in intimate relationships.

Ultimately, this thesis serves as a lodestar that is intended to guide designers through the inherent complexities of designing interfaces that leverage artificial agents to aid expression and understanding by the user. I aim for my research and exploration to apply to areas that extend beyond a focus on expression and understanding, and serve as a guide for anyone who is interested in learning how designers might address dimensions that cannot be easily measured.

Research Question

Over the course of my thesis, I was guided by the following design question.

How might artificial agents be designed to enhance intimate partners' capacity for expression and understanding in their relationship?

Overview of Study

My process was comprised of four phases: discovery, definition, exploratory, and generative. I designed each phase with the intent of deriving insights that could then inform work in future phases.

  • Discovery PhaseThe discovery phase centered around building a base of knowledge that could inspire and inform this thesis.
  • Definition PhaseThe definition phase focused on investigating the theory, potential applications, and affordances of relevant research areas by conducting literature reviews, artifact reviews, and conversations with experts.
  • Exploratory PhaseThe exploratory phase focused on studies designed to acquire a more comprehensive understanding of the models users have of artificial agents and intimate relationships. Emphasis was placed on developing insights based on first-hand experiences with an artificial agent and in an intimate relationship. These studies led me to design a set of 'How Might I...' statements that deeply informed concepts in the generative phase.
  • Generative PhaseThe generative phase was dedicated to creating and evaluating artificial agents designed to enhance an intimate partners' capacity for expression and understanding. Each artificial agent was designed to take a unique approach on how an agent could enhance intimate partners' capacity for expression and understanding.

Explore each of these stages below to see how they informed my work.

Prototyped Concepts

A large portion of the generative phase was dedicated to conducting investigations through prototypes. Explore each of the prototyped concepts below to see how they could enhance an intimate partners' capacity for expression and understanding.

Principles Derived

All phases enabled me to develop a number of principles to assist those who design for dimensions that make us human and cannot be easily measured. The following is a list of those principles gathered, framed, and written for a designer working specifically within the context of an intimate relationship and designing for expression and understanding. By providing these principles I believe designers can recognize deficiencies and opportunities within their design process.

  • Effectively IntegrateIntegrate and frame interfaces so that the intimate relationship is at the center of the interaction.
  • Allow For DiversitySupport a diversity of couples (i.e., those involved in very different practices, those that employ a variety of communication methods, those at very different stages of their relationship).
  • Design For SupportSupport rather than direct a couple (i.e., push couples to reflect on their relationship from different perspectives).
  • Evolve Over TimeEnable an experience to evolve as intimate partners' interact more with an agent (i.e., provide support customized to that couple, employ learnings from previous experiences).
  • Think Through ControlsConsider the tradeoffs of controls. For instance, if a user is allowed to choose specific metrics, could the design evoke greater bias in a relationship? What aspects of an experience is a user permitted to adjust?
  • Reveal the Good with the BadReveal both the positives and negatives of a relationship and its exchanges.
  • Create Moments of ContemplationEstablish touch points for reflection (i.e.,How can a person more effectively express themselves to their partner? How might an agent interpret those actions?) beyond an intimate partners' core interaction with an experience
  • Study Attitude Changes Towards the ArtificialUnderstand that an individual's attitude towards the artificial is constantly evolving (i.e., individuals have very different concerns about agents and individuals concerns over data privacy may change).

While it is important to consider the opportunities that each of these concepts provide, it is also important to look at the challenges each of these prototypes bring to light. For instance, many of these concepts are not possible with the current state of natural language processing and a considerable amount of additional work would need to be conducted to address the integration of these agents into a user's everyday life. The following is a catalog of those challenges, and theories on how they might be addressed.

  • Understanding the developing landscape of AI to ground designs in the state of the art and speculate future developmentsEven with the considerable developments in the field of artificial intelligence, sentiment analysis tasks are still limited to analyzing the polarity of a piece of text and dialogue tasks where a system tracks what a user wants from the system at each step, while communicating with that user still has approximately a 75% accuracy rate. Until sentiment analysis parses text beyond two dimensions, and the accuracy of dialogue tasks improves, additional frameworks will need to be developed that enable an agent to takea position similar to those found within the concepts I have explored.
  • Finding effective ways to consistently affect behavior changeWhen developing these concepts, I paid particular attention to the effect they could have when a couple interacts with them for the first and second time. While I considered how an agent's relationship with an individual or couple might evolve over time, further study is warranted to fully understand that evolution and enable an agent to positively affect that individual or couple in the long run.
  • Helping couples acknowledge the potential ways that they can improve their relationshipOne comment that seemed to pop up in some of my conversations with research participants was that a concept could really help other couples, but they had a hard time seeing how their relationship could benefit from such an intervention. Although this statement may be true, it may also confirm what a participant conveyed to me, when he said "we think we are good communicators when we are not". Bedtime Visions, Private Conversations, and my other concepts were attempts at challenging the notion that an individual or couple's current capacity for expression and understanding is fixed. Advancing these concepts and others so that individuals overcome the stigma associated with getting relationship help is a logical next step. The results of such work could provide individuals and couples who are not ready to seek out professional help with beneficial support.
  • Designing ways to consistently engage a user over timeWhen designing these concepts I paid particular attention to the first set of interactions an individual or couple might have in an experience. While those interactions may play a key role in engaging a user and highlighting the value of that experience, it may be no better than a self help book or an app if it does not consistently engage that user over a long period of time. An experience that evolves over time with a user could provide this value, making it both constructive and engaging.
Potential For Impact

As I began this project, I hypothesized that romantic relationships would provide a rich area to explore the relationship between humans and artificial agents, particularly because of the different forms of communication (verbal, non-verbal, etc...) employed, the high stakes at play, and the increased possibility for emotion to take the place of logic. Now at the close of the study, I believe that this thesis did not only explore that relationship but provides insight into how an artificial agent can:

  • Enhance an intimate partners' capacity for expression and understanding
  • Help a partner better understand themselves, their partner, and their relationship
  • Support a diversity of intimate relationships (i.e., intimate relationships can come in all shapes and sizes)

While also bringing to light:

  • The possibilities for artificial agents to successfully integrate into an intimate relationship
  • How design research can be applied to aid the understanding of the relationship between humans and artificial agents
  • Conversational symbiosis and how an experience can support it
  • How interfaces that integrate artificial agents can support conversational symbiosis while effectively integrating into an individual's life
  • How interfaces can support fluid exchanges between humans and artificial agents
  • How artificial agents can be well integrated into interfaces

By designing artificial agents, I was also able to explore ways of broadening an individual's perspective of relationships between humans and artificial agents through an expansion of approachable concepts. I recognize the need to create images that depict possibilities of relationships between humans and artificial agents that extend beyond those solely dealing with efficiency, effectiveness, and productivity. The artificial agents leverage images that use language accessible to a general audience. This approach strives to make concepts approachable, increase their appeal to people's imagination, and evolve their understanding of artificial agents.

As part of my thesis, I submitted a book designed to provide a detailed portrait of my work process. You can find that book here.