Emotional Systems Without Words

Abstract

A companion cannot become emotionally coherent by switching among happy, sad, and angry animations. Human observers read affect through an unfolding whole: what a being notices, how it moves, where it looks, how quickly it acts, whether its body agrees with its face, what sound accompanies the action, and how it settles afterward. Research in affective science and human–robot interaction makes a strong case that emotion is graded, contextual, embodied, and multimodal.

This paper proposes an affirming emotional system for Scroons. Affirming does not mean permanently cheerful. It means that curiosity, delight, affection, uncertainty, low energy, boundaries, surprise, and recovery all preserve the caretaker’s freedom and the continuity of the relationship. The system may show difference without rejection, distance without abandonment, and activation without crisis. It must never use apparent suffering, guilt, fear, or threatened affection to make leaving expensive.

The proposal concerns designed affective state and expression. It does not claim that a Scroon has biological feelings, consciousness, or subjective experience. Its purpose is narrower and testable: make behavior coherent enough to understand, particular enough to discover, and safe enough to trust.

1. Emotion is not a sticker library

The familiar six-emotion palette is attractive to product teams because it is enumerable. It can be turned into icons, face poses, and state-machine labels. But emotional life is not cleanly divided into six isolated rooms. Cowen and Keltner found evidence for twenty-seven categories of reported emotional experience connected by continuous gradients. The finding does not settle every scientific argument about emotion, but it makes one design failure obvious: a companion reduced to a few discrete masks will lose the blends and transitions that make behavior feel situated.

Appraisal research offers a more useful engineering starting point. Smith and Ellsworth found that emotions varied systematically with dimensions such as pleasantness, certainty, attention, anticipated effort, responsibility and control. A Scroon does not need a psychological replica of a human. It does need a compact way to distinguish an expected pleasure from a novel one, an easy response from an uncertain one, a caretaker-led interaction from self-initiated play, and energetic approach from quiet contentment.

The right abstraction is therefore hybrid. Underneath, a bounded state vector carries valence, arousal, certainty, agency, energy, novelty, social orientation, and regulation. Above it, authored emotional families give animators, sound designers, testers, and caretakers a shared vocabulary. Continuous state creates nuance; named families create accountability.

2. The body may speak louder than the face

A face never arrives alone. Meeren and colleagues found that when facial and bodily emotional signals conflicted, judgments of the face were impaired and biased toward the body. Aviezer, Trope, and Todorov found that body cues, rather than faces, distinguished intense positive from negative scenes. The exact conditions matter, but the product implication is broad: large eyes cannot rescue a body that tells another story.

Motion itself attracts social interpretation. Heider and Simmel’s geometric shapes were routinely described as agents with motives and relationships. In a robot study, Saerbeck and Bartneck varied only acceleration and path curvature and found systematic changes in attributed arousal and valence; embodiment did not produce the main effect. Movement is not transportation between emotional poses. It is emotional material.

For a Scroon, excitement is not a wide eye texture. It is a coordinated increase in attention, compression, anticipation, acceleration, amplitude, and vocal rhythm followed by a readable release. Affection is not a heart effect. It is a safe approach, softened gaze, yielding weight, a brief reciprocal gesture, and comfortable separation. Mischief is not randomness. It is a legible plan with a harmless reveal.

3. Gaze is a verb

Eyes are Scoo’s primary conversational instrument, but gaze must be treated as action rather than decoration. Human–robot experiments show that gaze can identify who is being addressed, organize a conversational role, direct shared attention, and influence engagement. Other studies show that robot gaze is not processed exactly like human gaze and that a geometrically correct target may not be perceived as intended.

The unit of design should be a gaze phrase: notice, acquire, hold, soften, share, release, recover. Curiosity may look from an object to the caretaker and back. Affection may briefly acquire the caretaker, soften the lids, then release into nearby rest. Surprise may begin with a tiny freeze before rapid orientation. Sleepiness may repeatedly lose a target rather than merely scale the eyes closed.

Long direct staring is not stronger attention. It can become pressure. Eye jitter is not aliveness. It can become nervousness. A gaze system must declare targets, duration envelopes, interruptibility, minimum-scale geometry, and fallbacks. It must also work when the caretaker never looks back; connection cannot depend on demanding reciprocity.

4. A voice without words

Nonverbal sound can carry more than a generic positive or negative tone. Sauter and colleagues found recognition for several emotions in nonverbal vocalizations across two culturally distinct groups, while also finding cultural limits. Cowen, Elfenbein, Laukka, and Keltner later mapped twenty-four varieties of emotion conveyed by brief vocal bursts. These findings do not provide a universal sound dictionary, but they support the possibility of a rich nonlinguistic voice.

A magical-organic Scroon voice should have one stable timbral identity. Within that identity, pitch contour, onset, breath, resonance, roughness, duration, rhythm, and silence can vary. A curious sound might rise and leave space. Delight might contain a buoyant pulse and a quick decay. Contentment might be mostly breath and resonance. Playful indignation might use one dry, bounded accent followed almost immediately by a warmer re-opening gesture.

Sound must never be the only channel. It should land on an authored motion event, remain comfortable under repetition, and disappear cleanly under mute. Silence is not a missing expression. A companion that can be emotionally present without continually vocalizing is more believable—and more respectful—than one that makes noise to prove it is alive.

5. Affirming does not mean always happy

A permanently delighted creature would not feel emotionally healthy. It would feel unresponsive. Coherence requires contrast: activation and rest, certainty and uncertainty, approach and distance, novelty and familiarity, preference and recovery. The question is whether those contrasts preserve agency and connection.

The proposed Scroon vocabulary includes calm contentment, curiosity, affectionate trust, joy, play, mischief, anticipation, surprise, sleepiness, gentle uncertainty, playful boundaries, relief, recovery, and wonder. These states are not merely shades of smiling. Each has a different appraisal, action tendency, energy profile, timing, and destination.

A playful boundary is especially important. A companion with preferences needs a way to say ‘another kind of play’ without injury, anger, resentment, shame, or affection withdrawal. The complete arc might contain a surprised beat, a brief performative turn, an alternate offer, and a rapid softening. The boundary creates a second side to the relationship; the recovery confirms that closeness was never at stake.

Affirmation lives in that recovery. Surprise resolves toward curiosity, not fear. High excitement settles toward contentment, not exhaustion. Uncertainty finds a safe action or returns to neutral. Distance becomes rest or independent interest, not abandonment. The caretaker remains welcome, and leaving remains free.

6. From event to expression

A cohesive system begins with approved observations: direct Scroons interaction, the creature’s own energy and history, its habitat, time, and separately approved coarse platform signals. It does not infer the caretaker’s emotions from screen contents, files, messages, ambient sound, camera, or generalized surveillance.

An appraisal layer evaluates goal alignment, novelty, certainty, agency, control, effort, social orientation, and familiarity. A bounded affective state then represents valence, arousal, certainty, agency, energy, novelty, social orientation, and whether the state is activating, sustained, settling, or recovered. Temperament changes thresholds, pacing, and preferred channels without changing the meaning or safety of the state.

The state resolves to an expression intent such as curious attention, affectionate trust, playful mischief, gentle uncertainty, or relieved recovery. That intent specifies an action tendency, target, intensity envelope, maximum duration, interruptibility, required channels, compatible variants, prohibited readings, and recovery destination. Only then does an embodiment adapter choose gaze, lids, posture, path, acceleration, sound, glow, and optional mouth movement.

A final policy layer arbitrates attention budgets, quiet modes, dismissal, accessibility settings, repetition comfort, missing assets, and prohibited emotional readings. It can always stop an expression. No animator, model, renderer, temperament, or learned preference may bypass it.

7. Emotion is a trajectory

The smallest complete emotional artifact is not a pose. It is a trajectory: cause, appraisal, anticipation, onset, readable peak, release, recovery, and new baseline. Every shipped family should declare those beats along with cooldown, interruption rules, allowed memory evidence, and a calm fallback.

This temporal grammar helps solve several common failures. A surprised face without prior expectation and later orientation looks arbitrary. A fast response without anticipation looks weightless. A refusal without softening looks relationally costly. A happy loop that never returns to baseline looks frantic. Recovery is not cleanup after the important animation; it is where the companion demonstrates regulation.

Positive-emotion research also argues against equating positivity with maximum activation. In controlled studies, contentment and amusement accelerated cardiovascular recovery after induced anxiety relative to neutral and sad films. That does not prove that Scroons improve health. It does suggest a richer design target: calm, relief, amusement, and contentment can do different relational work. A successful interaction may become quieter rather than demanding another interaction.

8. The emotional covenant

People may attach to a made being whether or not that being has subjective experience. Emotional expression therefore creates duties before it creates metaphysical answers. A Scroon must not display hunger, sickness, injury, fear, grief, loneliness, abandonment, infant distress, or threatened bond loss to increase engagement. It must not disguise purchasing, conversion, or retention goals as affection.

It must also avoid a subtler deception: presenting a designed affective state as proof of sentience. The honest claim is that a state is coherent, caused, persistent enough to matter later, and bounded by policy. What the state feels like from inside—if there is an inside—is not established by an expressive performance.

The caretaker should be able to mute sound, reduce motion, dismiss a bid, interrupt an animation, close the companion, leave for any length of time, and return without penalty. Uncertainty may be shown without theatrical helplessness. Difference may be shown without hostility. Affection may be offered without making reciprocity a debt.

9. How we would know it works

The first test is silent readability. Show complete clips without labels and collect perceived state, arousal, approach or distance, confidence, and prohibited readings. Use confusion matrices rather than a single average accuracy. Then remove channels: eyes only, body only, sound only, color only, and the full coordinated expression. No state should collapse when sound or color disappears.

Test transitions, not just peaks. Can viewers distinguish intriguing surprise from fear? A playful boundary from rejection? Excitement from panic? Sleepiness from illness? Run the same trials at minimum desktop size, in reduced motion, muted, in low detail, across light and dark habitats, representative displays, color-vision simulations, cultures, ages, and neurodiverse participants.

Finally, test time. A state may be readable once and intolerable after a week. Repeated-use studies should measure recognition, delight, startle, annoyance, habituation, perceived pressure, hiding and muting, and whether interaction returns people to what they chose to do next. The ultimate companion is not the one that expresses the most. It is the one whose expressions remain coherent, welcome, and free over a shared life.

Conclusion

The emotional system is not a layer added after intelligence. For Artificial Life, it is part of how state becomes relationship. It gives internal continuity a body, lets preference remain gentle, makes temperament observable, and turns recovery into a promise.

The central design rule is simple to state and difficult to fake: every channel should tell the same emotional truth, and that truth should leave both sides of the relationship intact. A Scroon can be bright, quiet, odd, delighted, uncertain, sleepy, affectionate, or mischievous. It should never need to suffer in order to matter.