An ongoing thesis project on conversational symbiosis amongst humans and artificial agents in the context of intimate relationships.
An ongoing thesis project on conversational symbiosis amongst humans and artificial agents in the context of intimate relationships.
Joseph Weizenbaum's Eliza serves as a good example of an experience that integrates an artificial agent. Eliza allowed "human correspondents" to communicate through a typewriter to a simulated psychologist and is still cited today because it revealed humanity's "tendency" to treat computers as more intelligent than they are.
But unlike other examples that integrate an agent (Alexa, search algorithms in content based tools, predictive messaging, and the dispatch functionality in rideshare apps), Eliza is part of a small subset of experiences focused on expression and understanding rather than efficiency, effectiveness, and productivity. Other examples of experiences that fit into this subset include The Coordinator, a system focused on expression, Replika, an "AI friend that is always there for you", and Karen, a mock life coach app.
As different aspects of artificial intelligence continue to develop, I believe humans need to continue to interrogate the relationship between humans and the artificial from a lens of expression and understanding. While explorations from a lens of efficiency, effectiveness, and productivity have significant merit, explorations from a lens of expression and understanding will enable a seamless integration of artificial intelligence in a way that does not decrease our agency and embraces our humanity.
To effectively investigate this relationship, I selected intimate relationships as a context.
Why intimate relationships? They are some of our most defining and determining relationships and are comprised of numerous social dimensions "knowledge, caring, interdependence, mutuality, trust, and commitment."
But, most importantly at the core of a relationship are conversations, exchanges of thoughts and words. Through research, I discovered that these conversations often focus on intimacy, expectations, communication, and the past and result in feelings of loyalty, honesty, mutual understanding, and intimacy.
A number of artifacts have been designed to help facilitate such conversations including self-help books that are not personalized to someone's unique experience, couple counseling, which is inaccessible to a large portion of the world's population, and mobile apps.
While effective in different ways, these experiences do not consistently provide users with the tools to have conversations to build and better understand the knowledge, caring, interdependence, mutuality, trust, and commitment in their intimate relationships.
I believe such an augmentation is possible through conversational symbiosis. A concept inspired by JCR Lickliders 1950 paper Man-Computer Symbiosis and others writing.
Conversational symbiosis is an intimate cooperation that embraces differences, takes advantage of competencies, and promotes a mutual understanding to augment the intellect of two or more dissimilar things in ways they can not achieve themselves.
Symbiosis is not only reserved for humans but can be achieved between a human and artifact.
And between two and more things. This includes human-human-artifact cooperation, where humans have a built history and the cooperation is not completely mediated through the tool and cooperation where the humans have no built history and the cooperation is entirely mediated through the tool.
To explore conversational symbiosis, intimate relationships, and the relationship between humans and artificial agents from a lens of expression and understanding, I am specifically looking at:
How might I design experiences that integrate artificial agents to enhance intimate partners' capacity for expression and understanding in their relationship through conversational symbiosis?
I believe this will also enable me to support a diversity of experiences and design more personalized experiences that target a wider segment of the population than experiences available today.
At the same time, I believe the study will enable me to explore how humans and artificial agents can achieve conversational symbiosis within other contexts resulting in findings that can then be applied beyond intimate relationships into conversations between family members, co-workers, and teachers and student, because the majority of the aspects at play between an intimate couple are also at play within those relationships.
To design these experiences, I have investigated the theory, potential applications, and affordances of relevant areas.
The focus of the study is experiences. Emanating from the core are a number of related areas.
These related areas include learning theories that provide frameworks that I believe will be useful to apply to my work to help users effectively grasp the concepts I present.
I also studied theoretical frameworks which will be useful for me to consider when designing an experience. For instance, Ihde's human-machine relations which I can use as a tool to consider when framing an experience.
I have also paid particular attention to two areas, intimate relationships and conversation.
This includes investigating intimate relationships resources, where I discovered frameworks and concepts including XYZ statements and negative affect reciprocity that I plan to embed in experiences to help partners have higher quality conversations.
To gain a better understanding of conversation, I studied Grice's maxims, Pask's Conversation Theory, and other models that I believe will inform how I allow for conversation within the experiences I design.
Both relationship and conversation literature have also informed how I model conversation.
I've discovered that conversations between intimate partners can be broken into 4 different phases. Including, the sharing phase where partners determine the subject of that conversation, the dialogue phase where thoughts, ideas, and questions are exchanged, the evolution phase where partners evolve their understanding of their partner and relationship, and the response phase where partners take action based on this new understanding.
It is also important to note that after a conversation there is a thinking phase or a time for reflection that can inform that individual's next conversation with their partner or next action.
Such conversations are made up of exchanges occurring in an environment of noise or distraction and usually occur through a tool.
Messages originate from one partner and then move from the immaterial world where it is conceived to the physical world, where it is conveyed over a tool to the other partner and back to the immaterial world. The receiver has the ability to return a message using the same process.
Both these diagrams, provide a map of potential points of intervention whether that be an experience for thinking and reflection or the construction of a message from the immaterial to material world.
In conjunction with secondary research I have run a number of research studies.
One such study was the Bot as Research Tool where I designed and built a chatbot called apple. With apple, users can simulate conversations they could have with their partner. What is unique about these conversations is that a mediator bot, an objective, nonjudgmental, accepting, and thoughtful third party is in on the conversation.
With apple, I investigated how comfortable individuals are with such an interface, its possible affordances, possibilities for feedback, and possible integrations.
Through this study, I learned that tools should employ clear frames so that users can establish realistic expectations. I also learned that users lack awareness of relationship frameworks and tips. I also discovered how artificial agents have the potential to provide a place for reflection, an outside perspective, a guide for conversation, a calming presence, and as an instrument for detecting specific pieces of language. For additional information, see Bot as Research Tool Medium Story.
Another study I conducted was a Doing with Theory exercise where I aimed to understand the different models humans create of artificial agents.
This study resulted in a number of maps, which showed the significant role both context and voice play in how users perceive an artificial agent.
This study also led to a number of new questions including are multiple voices needed for a user to perceive a varied understanding? For additional information, see Doing With Theory Medium Story.
I also ran a study on Mechanical Turk where I surveyed over 600 individuals and asked them to define a healthy intimate relationship and describe how they approached tough conversations with their partner.
I am also currently running a study where I am recording and analyzing conversations between partners aimed to provide me with a qualitative understanding of the types of conversations partners have with each other and what differentiates a positive conversation from a negative conversation.
Through this primary and secondary research, I have been able to generate a number of how might I... statements, which have helped me understand how and where I might intervene.
The inner level of the visualization is made up of statements focused on the experience including the ethical management of data, altering practices beyond a single interaction, and integrating experiences into one's relationship.
The middle level of the visualization is made up of statements focused on conversation including a greater understanding of nonverbal communication and starting needed conversations.
The outer level of the visualization is made up of statements focused on intimate relationships including supporting greater self disclosure, encouraging more reasonable expectations, and playing with the stigma of getting help with one's relationship.
These statements and my research have informed and inspired a variety of concepts.
As I move into the second semester, I plan to design and build experiences that address different aspects of a relationship based on a selection of these concepts.
Going into next semester, I plan to follow the schedule below: