LLenny's Podcast
← All frameworks
InnovationCaitlin Kalinowski (ex–OpenAI, Meta, Apple)

Designing Robots That Feel Non-Threatening

Make robots soft, attentive, and telegraph intent before they move.

Difficulty
Moderate
Time to result
~months to results
Steps
3
Confidence
88%

Humans carry deep expectations about how another being should behave when it enters a shared space, and robots that violate them read as creepy. Kalinowski's design principles — learned from HRI researcher Leila Takayama — are to make robots appear soft and reactive, signal awareness of the person, and, critically, telegraph intent before acting (look before you turn). Animation studios like Pixar and Disney are the gold standard for conveying intent and approachability.

Origin

Attributed to human-robot-interaction researcher Leila Takayama, who taught Kalinowski these principles; Kalinowski also credits Pixar and Disney as best-in-class at this design work.

Core principles

  • 01Humans expect other beings to acknowledge them when entering a space
  • 02Robots should appear soft, reactive, and attentive to the person
  • 03Telegraphing intent before acting removes the startle that makes robots creepy
  • 04Character animators (Pixar/Disney) are world-class at conveying emotion, intent, and approachability

How to run it

  1. 1

    Make the robot appear soft and reactive

    Design for a non-threatening presence: soft appearance, visible reactivity to the person, and cues that the robot knows you're there and is attentive to helping you. This is both a safety property (softer, lower-mass parts) and a social one.

    Watch out With notable exceptions, many current humanoids are surprisingly creepy — it's easy to default to threatening.

  2. 2

    Signal awareness when someone enters or exits

    Replicate the nonverbal acknowledgment humans give each other — you don't have to talk, but you look up. A robot that ignores a person entering the room breaks the expectation and unsettles them.

  3. 3

    Telegraph intent before acting

    Show intent physically before executing a motion. A robot that suddenly turns and does something scares people; a robot that looks before it turns, then goes, is far less alarming. Build this 'look-then-act' cue into any robot, humanoid or not.

    Pro tip Study character animation — Pixar and Disney are the world's best at showing intent, engagement, and approachability, even without doing it at physical volume.

In the wild

Look-before-you-turn

Kalinowski contrasts two behaviors: a robot that suddenly turns and acts, which scares the person, versus a robot that first looks in the direction it will move and then goes, which reads as much less alarming — the same physical action, made safe-feeling by telegraphing intent.

The intent-signaling version preserves human comfort and trust in shared space; the abrupt version reads as creepy or threatening.

Common mistakes

Moving abruptly without signaling intent

A robot that suddenly turns or acts startles people and erodes the trust needed for it to operate near humans; the fix is cheap — look before you move.

Ignoring social acknowledgment cues

A robot that doesn't register a person entering or leaving violates a deep human expectation and lands as unsettling, no matter how capable it is.

Is it for you?

Best for

Robotics and product teams designing machines that will share physical space with people at home or work

Not ideal for

Caged industrial robots or non-embodied systems where no human shares the robot's immediate space

From the transcript

her name's Leila Takyama

1:06:30

if a robot just suddenly turns and does all this stuff, it scares you. But if a robot looks before it turns and then goes,…

1:08:00

Pixar, Disney are probably the world's best at doing this type of design work.

01:00

From the episode

Why we’re at the beginning of the AI hardware boom

Caitlin Kalinowski (ex–OpenAI, Meta, Apple)