![[L - R: Film Critic Manuel Betancourt and Dr. Jonathan Gratch [photo taken by USC PhD student Bin Han]](https://ict.usc.edu/wp-content/uploads/2025/05/Gratch2.webp "[L – R: Film Critic Manuel Betancourt and Dr. Jonathan Gratch [photo taken by USC PhD student Bin Han]")
[L – R: Film Critic Manuel Betancourt and Dr. Jonathan Gratch [photo taken by USC PhD student Bin Han]
By Dr. Jonathan Gratch, Director, Affective Computing, USC ICT
On May 11, 2025, I participated in a panel at the Sloan Film Summit, held at the Academy Museum of Motion Pictures. The focus of our discussion was the film Love Me, directed by Sam and Andy Zuchero, featuring Kristen Stewart and Steven Yeun. The film presents a speculative narrative in which two artificial intelligences, long after the extinction of humanity, reconstruct human concepts—most notably love—by watching billions of YouTube videos and engaging in repeated virtual interactions. Over time, these entities evolve into something recognisably human.
The Sloan Foundation recognised Love Me for its engagement with scientific themes. My role on the panel was to speak to the plausibility and implications of the film’s premise through the lens of contemporary affective computing research. Though I touched only briefly on my own published work [including: These aren’t the droids you are looking for: Promises and challenges for the intersection of affective science and robotics/AI] the film’s narrative aligns strikingly with several ongoing lines of inquiry in AI.
The scenario portrayed in Love Me—training on the remnants of human digital culture followed by iterative self-interaction in simulation—closely parallels how large language models are developed. In one of my collaborations with Dr. Gale Lucas, Director, Technology Evaluation Lab, USC ICT [Be selfish, but wisely: Investigating the impact of agent personality in mixed-motive human-agent interactions] we explored this learning framework in the context of negotiation: artificial agents trained on broad human data, then refined through situated interactions. The resulting behaviors suggest that with appropriate training, machines can develop complex social strategies. While not love in any meaningful psychological sense, these behaviors indicate a capacity for adaptive, socially aware action.
The film also addresses a conceptual issue in AI: the problem of recursive stagnation. The characters—disembodied AIs—become caught in repetitive simulation loops. Only through the emergence of emotion are they able to escape these cycles. This narrative device echoes a long-standing idea in cognitive science—that emotion is not epiphenomenal, but rather essential to intelligent behavior. I referenced an early but foundational paper, Motivational and emotional controls of cognition, which argues that emotion plays a critical role in interrupting unproductive cognitive routines and promoting behavioral flexibility. Related neuroscientific work by Antonio Damasio and Antoine Bechara at USC supports this view, demonstrating that affective signals are integral to decision-making processes in humans.
I briefly mentioned one of my own early contributions to this area, A step toward irrationality: using emotion to change belief, which examined how emotional interventions can alter belief structures—essentially demonstrating that emotion can override purely rational inference when circumstances demand it.
During the audience discussion, ethical considerations emerged. What happens when artificial agents become sufficiently socially competent that people prefer their company? In our study on “social snacking” [Social snacking with a virtual agent–On the interrelation of need to belong and effects of social responsiveness when interacting with artificial entities] Dr. Gale Lucas and I showed that interacting with a socially responsive AI can temporarily satisfy one’s need to belong—sometimes at the expense of seeking real human contact. These findings raise non trivial concerns about the displacement of human relationships by artificial companions.
Finally, I addressed the film’s implicit claim that emotion is the final frontier separating humans from machines. This idea—that what distinguishes us from AI is our capacity for feeling—is in fact a relatively recent cultural shift. Historically, emotion was regarded as a remnant of our animal heritage, with reason being the defining trait of humanity. As machines begin to match or exceed human reasoning, we are perhaps reconfiguring our identity to preserve a sense of uniqueness. I referenced a paper (not my own) on how humans respond creatively when their distinctiveness is threatened by comparisons with machines. Emotion becomes a last bastion of human exceptionalism.
Love Me does not offer answers, but it does pose rigorous questions. What role should emotion play in artificial cognition? Can artificial entities develop emotionally functional behavior without biological substrates? What are the social consequences of widespread emotional machines? These are no longer science fiction questions—they are active research concerns. And as we move forward, films like Love Me serve not just as cultural reflections, but as provocations to scientific inquiry.
//