AI — the technological Mirror
Most tools invented in history are a reflection of needs and desires. Throughout history our tools improved and often the advances of specialized tools have been the culmination of collective wisdom materialized, making profound changes to our societies.
With the advance of self-learning artificial intelligence (AI), like Machine Learning, Neural Nets and similar, we seem to move towards another cusp between what technology can do and the effect it has on society. We are teaching our tools to understand the world in more human ways by making them realize the patterns and relations that govern our everyday.
The reflection on human behavior is an inherent aspect of AI development. Either through the data it analyzes, or the way it looks at and uses data, reflecting on individual or collective human behavior is part of AI.
Machine learning´s “magic” is found in that the software itself searches for rules and models to connect input and output, leaving room for interpretation by the machine so to say. Most of the time this lack of initial definition is surpassed by the amount of data a system can search and learn from. Nevertheless, there are multiple aspects of AI and user involvement that are being explored ranging from how AI is trained with the help of humans, to interactive AI applications that use and convert human input, to human augmentation with AI.
In the field of Interactive machine learning, studies have already produced self-reflective behavior on the human side, even when it was not an intended goal. Rebecca Fiebrink, the creator of the Wekinator, a tool that aims to make IML more accessible for creative creators like musicians and artists, revealed that in her study the participating professional cellists gained “a new perspective on her own bowing technique, when occasionally she discovered through consistently poor model performance that her training data was not as clear as she thought it had been”.[1]
Through this reflective nature of AI, Design and designers are able to naturally look at the human element engaged with the technology. And it offers a chance to imagine new applications for users.
Replika.ai, is one such commercial product that is using the idea of reflective AI agents. This application is a mix of an interactive diary with the promise of creating an AI agent in the form of a chatbot based on the user’s writings. The user based chatbot would then be a technological mirror for the person and act as discussion partner. Although the overall benefit of the app currently is closer to a digital diary that helps people to reflect on what is happening in their life, the project originated from Eugenia Kuyda´s desire to bring the memory of a passed friend to life. Originally, the team took the digital fragments that we all leave with other people each day in form of text messages, notes and e-mails and trained an AI on it. In that way giving friends a chance to speak to the lost friend. A situation that was until now only found in sci-fi like the very similar Black Mirror episode “Be right Back”.
On a happier note, creators are putting the idea to use in many more ways, games such as ECHO by Ultra Ultra, are utilizing the power of combining the established concept of game AI with new game mechanics. In ECHO, the game uses the idea of reflective AI agents literally, creating NPC´s that are copying the players behavior that turn against the player.
Of course, it´s still our human needs and desires that drive development — but to utilize these new generation of tools, we need to look at ourselves and how we understand the world around us. It will be necessary to determine what behavior, knowledge, believes or similar can be learned by AI and what are the necessities to do so, while determining feasibility and real user benefits.
Looking at the human center of any such application, Self-reflection or introspection, is most of the time a natural process in our lives, something we do more or less consciously, but it can be trained like a technique and used just as a powerful tool.
John Dewey defined reflection in his book “How we think” in 1933 as “active, persistent and careful consideration of any belief or supposed form of knowledge in the light of the grounds that support it and the further conclusion to which it tends’’.[2] He continues to state that reflection happens when “a person finds himself confronted with a given, present situation from which he has to arrive at, or conclude to, something that is not present. … What is present carries or bears the mind over to the idea and ultimately the acceptance of something else.”
The interest in reflection for practitioners increased because of institutional demand for professionals to document and show that continuous efforts of education are made to ensure that their practice is up to current standards. Which is especially necessary for professions where ongoing progress changes processes for the practitioners- as is the case with healthcare and education itself. Especially in professional development we can find practices and processes that are meant for efficient application and might favor an adoption and symbiosis with AI technology.
I believe self-reflection and self-development are avenues that could benefit from these new developments, a viewpoint that others such as replika.ai seem to share. At the moment, a lot of research is going into eliminating the human bias in these new technologies - trying to make them as objective as possible. But at the same time, the things that define us as individuals are worthy to be investigated too. AI applications have already shown their capability in turning pictures into art-like paintings by your favorite painters by being trained on their one particular style. Building on that, we can imagine that humans will train their personal AI´s to be reflections and extensions of themselves, elevating what that person can do, from completing laborious tasks in a certain way, to facilitating communication with other AI´s and their humans on a much grander scale than what was possible before.
By training an AI on a human, a digital agent can be created that inherits an individual’s traits, representing a specific facet of that person. That makes self-reflection, augmentation/collaboration and new expression possible for that person.
How can users benefit from this? What form should these agents take? These are the questions I want to inquire about in my ongoing master thesis project.
Please reach out to me if you have something to add — ideas, references or are just curious. Let me know.
[1] “Real-time Human Interaction with Supervised Learning Algorithms for Music Composition and Performance”; Rebecca Fiebrink; PhD Dissertation (2011)
[2] „How we think”; Dewey (1933); p.118
