Thinking in Pictures: Visual Analysis of Digital Owl Behavior

by Temple Grandin, Ph.D.
Professor of Animal Science

When I first looked at this owl simulation, I didn't see code—I saw a movie playing in my head. Twenty-four owls flying in three-dimensional space, diving and climbing, their energy bars depleting like fuel gauges. I can see it the way an owl sees it: the world from above, mice scurrying below like moving dots on a radar screen.

The Pressure of Hunger: Understanding Energy Depletion

The most striking thing about these digital owls is how their behavior changes with energy levels. When energy drops below 20, they're forced to rest. This is exactly like real animals experiencing hunger stress.

STRESS POINT: Energy < 20 = Forced Rest
In real owls, this would be the point where the bird's body starts shutting down non-essential functions. The simulation captures this biological imperative perfectly, even if accidentally.

I see this in cattle all the time. When cows get too hungry, they stop moving efficiently. They stand still more. They conserve energy. These digital owls do the same thing—they fly up to 150m altitude and essentially "park" themselves.

ALTITUDE AS EMOTIONAL STATE

200m ━━━━━━━━━━ SAFE/RESTING (Like cows in the back of the pasture)
150m ━━━━━━━━━━ RECOVERING
100m ━━━━━━━━━━ ALERT/WATCHING
50m ━━━━━━━━━━ ENGAGED/INTERESTED
0m ━━━━━━━━━━ HUNTING/STRESSED (Like cows at the feed trough)

The Problem with Perfect Vision

Every owl in this simulation has perfect vision within its range. This is completely wrong from an animal behavior perspective. Real owls have:

I'm seeing each owl's vision cone in my mind like a flashlight beam. But real owl vision is more like a flashlight with dying batteries—bright in the center, dim at the edges, with dark spots where the beam doesn't reach.

The Missing Fear Factor

MAJOR FLAW: These mice show no individual fear response!

Real mice would have varying levels of fear. Some would be bold (venturing far from cover), others timid (staying near edges). The simulation treats all mice identically, which creates unrealistic clustering patterns.

CATTLE HANDLING PARALLEL:
When designing cattle facilities, I always account for the "bold" cows who lead and the "scared" cows who follow. A good system works for both personality types. This simulation only models one personality type for mice—that's like designing a cattle chute for only the calmest cows.

Flight Patterns and Movement Quality

The way these owls move bothers me. Look at this code:

// Random wander
this.acceleration.x += (Math.random() - 0.5) * 0.1;
this.acceleration.y += (Math.random() - 0.5) * 0.1;
this.acceleration.z += (Math.random() - 0.5) * 0.05;

This creates jittery, nervous movement! Real owls glide smoothly. They don't jitter randomly. They make deliberate movements.

BETTER APPROACH: Use sine waves for gentle banking, smooth curves for patrol patterns. Real owls fly in arcs and circles, not random jerks.

The 3AM Mouse Convention: Emergent Prey Behavior

This is the most fascinating part! When predator pressure drops (most owls resting), mice naturally congregate. This is EXACTLY what happens in nature:

Real-World Parallel: Deer coming to water holes at dawn when predators are less active. Cattle moving to open pasture at night when wolves historically hunted less. The simulation accidentally discovered a fundamental prey behavior!

I can see it like a time-lapse movie: mice spreading out when owls are active, condensing into groups when safe. It's like watching iron filings respond to magnets—pure physics expressing as behavior.

Welfare Concerns for Digital Animals

I know they're just code, but the energy system creates genuine welfare issues:

DIGITAL SUFFERING INDICATORS:

If these were real animals, I'd redesign their environment immediately. Even in simulation, we should model humane conditions.

Visual Thinking vs. Code Thinking

How Coders See It:
- O(n²) complexity
- 60 FPS update loop
- State machine transitions
- Memory allocation patterns
How I See It:
- Owls circling in holding patterns
- Energy bars depleting like gas gauges
- Mice flowing like water away from danger
- Altitude creating emotional zones

Practical Improvements from Animal Behavior

1. ADD INDIVIDUAL DIFFERENCES:
- Some owls better at hunting (70-90% success)
- Some mice more vigilant (detect predators earlier)
- Personality types: bold vs. cautious
2. REALISTIC VISION:
- Reduced accuracy at distance
- Head movement requirements
- Weather affecting visibility
3. PROPER REST BEHAVIORS:
- Preferred perch locations
- Social roosting for some owls
- Territory defense even while resting
4. FEAR GRADIENTS:
- Mice avoid areas of recent kills
- Safe zones develop naturally
- Escape routes become established

The Beauty of Emergent Behavior

Despite its flaws, this simulation achieves something remarkable: complex behavior from simple rules. It's like my cattle handling designs—curved chutes work because they match natural behavior, not because they're complicated.

"The best designs aren't the most complex ones. They're the ones that work with natural behavior patterns, not against them."

When I close my eyes, I see the entire ecosystem running like a perfectly choreographed dance. Each owl traces its path through 3D space, leaving trails like ribbons. The mice flow below like schools of fish. It's beautiful in its simplicity, even with all its biological inaccuracies.

Final Thoughts

This simulation makes me think about how we model animal behavior in our minds versus in code. Programmers think in algorithms and data structures. I think in movies and sensory experiences. Real animals live in a world of sensations, fears, and drives.

The fact that simple rules create the "3AM Mouse Convention" gives me hope that we can model more complex behaviors. But we need to think like animals, not like computers.

REMEMBER: You can't design a good cattle handling facility by thinking like an engineer. You have to think like a cow. Same principle applies here—to model owls, think like an owl.

"Nature is cruel, but we don't have to be."
Even in our simulations.