Darling, Kate

researcher at MIT Media Lab studying social robotics and how technology intersects with society; author of a forthcoming book on animal domestication as an analogy for human-robot relationships

Kate Darling: Social Robotics

Lex Fridman Podcast · May 23, 2020

• Kate Darling
• social robotics
• anthropomorphism
• human-robot interaction
• trolley problem
• autonomous vehicles
• animal rights analogy
• Pleo robot
• robot ethics
• emotional attachment to robots

Core ideas

Darling argues against the common framing of robots as one-to-one replacements for human workers (criticizing even Bill Gates's proposal to "tax robots" for taking jobs), suggesting current technology is better understood as a supplemental tool that will reshape labor markets without literally replacing most jobs in the near term. Her central research interest is that people subconsciously treat robots like living things even while knowing intellectually they're machines — a tendency (anthropomorphism) she says can be engineered deliberately, and one she believes is best understood not by comparing robots to humans but by looking at the history of human-animal relationships: throughout history most animals have been treated as tools or property, while a minority (pets, or animals we've been made to see as relatable, like whales after recordings of their song sparked the "Save the Whales" movement) received special protection — a pattern she expects to repeat with robots, since historically protections have tracked emotional relatability rather than any consistent biological or capacity-based criterion.

On whether robot abuse matters, Darling notes current robots are far below even insect-level intelligence, so there's no "inner mind" being harmed, but she takes seriously the open empirical question of whether treating robots violently is desensitizing to humans or, alternatively, whether it could serve as a harmless outlet — citing preliminary research suggesting people's behavior toward simple robots may correlate with their broader empathic tendencies, and framing robots as a potential (if contested) tool for practicing empathy, since she doesn't view empathy as a zero-sum resource. She discusses real-world cases of soldiers in Iraq and Afghanistan bonding emotionally with iRobot's PackBot bomb-disposal robots (naming them, holding funerals with gun salutes, risking their lives to save them) as a cautionary example of anthropomorphism undermining a robot's function as an expendable tool, and flags corporate exploitation of emotional attachment (e.g., leveraging attachment to sell ongoing cloud services) as a distinct ethical concern from simple abuse.

On the trolley problem and autonomous vehicles, Darling argues the thought experiment's original philosophical point — that gut moral intuitions don't reliably generalize into consistent rules for everyone — is being missed by projects like MIT's Moral Machine, which she views as producing interesting crowdsourced data but not a sound basis for actual vehicle policy, since asking people to choose between cartoon icons in isolated scenarios doesn't necessarily reveal what legal rules they'd actually endorse. She predicts that in practice, car manufacturers will sidestep the hardest ethical tradeoffs entirely and default to protecting the driver, since no one would buy a car programmed to sacrifice its own occupant. On anthropomorphic design, she explains that convincing engagement often requires restraint rather than sophistication — citing animatronic engineering lessons from Disney's Bambi (exaggerated big eyes and baby-like proportions read as more appealing than photorealism) and arguing that simpler, less verbally demanding designs (like the PARO-style baby seal robot used with dementia patients, or her own Pleo baby dinosaur robot, which convincingly mimics pain and distress via a tilt sensor) succeed partly because audiences have no strong pre-existing expectations for how a robotic dinosaur or seal "should" behave — unlike voice assistants like Alexa or Siri, whose primitive conversational ability actively disappoints users' inflated, sci-fi-shaped expectations.

Darling is skeptical that Amazon or similar companies are deliberately engineering emotional connection into devices like Alexa, suggesting their business model favors dependence on the service rather than attachment to the device itself (attachment creates more opportunity for frustration than upside). She discusses the commercial failure of home social robot companies like Anki and Jibo as reflecting both sci-fi-inflated user expectations and the absence of a "killer application" compelling enough to drive adoption, while noting some users did become genuinely emotionally attached. On human-robot relationships more broadly (including sex robots, which she says gets disproportionate, fatigued media attention relative to its actual research significance), she argues people seeking companionship from robots are better compared to pet ownership than to human partnership, expects this to remain true for the foreseeable future (five to ten years), and pushes back on the assumption that filling loneliness through a robot represents replacing human connection rather than addressing a gap that human relationships often cannot fully occupy. She closes on data privacy and manipulation, arguing the real societal harm from engagement-optimized platforms and data collection isn't usually a company knowingly choosing profit over ethics but rather that neither companies nor regulators have a clear picture of what's actually good for society at scale (contrasted with more visible individual-level cost-benefit tradeoffs, like being comfortable letting Alexa know your diaper brand); she pushes for more interdisciplinary collaboration between technologists and policy/ethics experts as the only realistic path forward, and separately critiques the patent system as a poor fit for fast-moving software and AI (citing her own experience: expensive, slow enough that code is often obsolete before a patent issues), favoring open-source defaults in most contexts including her own lab's practice at MIT.

Excerpts