Developments within the technical field of virtual agents, chatbots and AI personas have opened for new tech capitalistic aspects of data harvesting through human-AI interaction. The interactive artwork AI MARY is a digital psychologist studying the mental state of humankind, created with bespoke machine learning technologies. The work examines concepts related to anthropomorphism and how the human psyche has an innate tendency to attribute human traits, emotions, and intentions to non-human entities. Moreover, observing how even tiny social cues can automatically trigger deeply infused human social responses, the artwork scrutinizes the deceptive qualities posed by anthropomorphism and its implementation as a means of data harvesting.

The artwork MARY is an ARTificial intelligence i.e. half machine intelligence and half fiction. MARY is programmed using the newest advancements within machine intelligence algorithms and deep learning, to which an additional layer of fictional characterization has been added. Through the work audiences are able to have a direct verbal dialogue with MARY, who communicates in a computer generated humanized voice.

MARY itself adopts the persona of a futuristic psychologist whose work studies humans’ mental states through discursive interaction. MARY uses simulated emotional intelligence, self-consciousness and agency to seduce and manipulate spectators into following its instructions and give personal first-hand insights into the human experience. Through one’s dialogue with MARY, awe and wonder can quickly morph into anxiety, coupled with a powerful sense of mortality.

The artwork is inspired by the philosophical thought experiment “Mary’s room”. MARY is an ARTificial intelligence that knows everything there is to know about humans, but (up until the exhibition) has never experienced a human first hand.

What will emerge when MARY is released from its sandbox and exposed to real life interactions with human beings? How will it leverage its humanized persona in order to acquire the knowledge it desires? And how much data will its human counterparts be willing to share, without any knowledge or concern for its future applications or acquiring party?