Apple Interaction

Apple Interaction is a virtual reality environment set in a circular woodland meadow, with a single large tree at its centre and a squirrel non-playable character (NPC) as the participant's counterpart. It was designed to elicit PLAY. Accordingly, it contains no score, visible timer, or explicit objective, and each session lasts 7.5 minutes.

The interaction centres on throwing and catching apples. Apples rest on the ground and grow on the tree, and the participant can pick them up and throw them to the squirrel. The squirrel monitors the participant's position and the trajectory of thrown apples, and responds by chasing and catching them, offering an apple to the participant, dodging a throw, evading in a keep-away pattern, or climbing the tree to drop an apple from above. Either party can initiate an exchange, producing a reciprocal sequence of turns.

The environment is intended to make this kind of interaction feel natural without imposed rules. It combines an open, daylit setting with a responsive and expressive squirrel, so that play can arise from the interaction itself.

The squirrel

The squirrel is a navigation-mesh agent controlled by a finite state machine, in which it occupies one behaviour at a time (for example idling, offering an apple, chasing, or climbing) and transitions between behaviours in response to events. A lightweight mood model tracks two variables, playfulness and tiredness, which scale its movement speed and bias its choice of behaviour over the course of a session. This mood model is the only adaptive logic in the environment; there is no difficulty adjustment or scoring.

The squirrel's sensors register the participant's distance and idleness, together with the trajectory and force of thrown apples. Its responses to being thrown at include catching, dodging, a startle, a playful trip, and a taunt, and its catch animation varies with the height and side from which an apple arrives. Its facial expressions are hand-drawn textures that change with mood, gaze direction, and blinking, and the pitch and volume of its vocalisations vary with its mood state.

The meadow and the player's actions

The scene is a circular woodland meadow centred on a large tree. Up to five loose apples rest on the ground and respawn at fixed points when fewer than three remain; further apples grow on the tree and become throwable when grabbed. Apples use a custom shader with a star-shaped sparkle effect and are available in several colours, and the grass uses a separate custom shader. The participant moves using short comfort hops in the direction of gaze, and grabs and throws objects with the controller trigger.

Design rationale

An animal NPC was chosen deliberately. Prior work on virtual companions indicates that a stylised, non-human character supports social presence and positive affect more effectively than a realistic human avatar, and studies of virtual pets report that responsive animals, including ones that bring objects to the user, help to initiate and sustain engagement. A stylised animal also avoids the uncanny valley, the unease that near-realistic, humanlike characters can provoke when they are almost, but not quite, convincing. The squirrel was therefore given a deliberately cute, stylised appearance, to encourage approach and rapport without that risk. Its mood-dependent facial expressions and vocalisations serve as social cues, since expressive behaviour, gaze, and vocal intonation are known to make virtual agents more believable. Soft, rounded textures were chosen because they have been associated with warmth and joy.

The remaining sensory choices follow similar evidence. Natural, open, daylit environments have been linked to more positive affect than urban ones, bright colours are associated with joy, and faster-tempo music is generally judged to be happier, so a moderately fast musical loop and ambient outdoor sound were used. Locomotion uses short hops to reduce the risk of cybersickness. No goal is imposed, in keeping with the autotelic nature of play.

Gallery

The procedure shared by all three environments is described on the Method page. For Apple Interaction, 25 participants took part in a within-subjects design with three measurement points: P1 before any VR exposure, P2 after the neutral tutorial scene, and P3 after the Apple Interaction environment. The Self-Assessment Manikin (SAM), PANAS-X and the Affective Neuroscience State Scales (ANSS) were completed at all three points. Presence (iPQ) was measured at P2 and P3, co-presence and social presence at P3 only, and behavioural telemetry was logged throughout the session.

Each part of the definition of PLAY was mapped to a measure. Valence, arousal and dominance were measured with SAM, joy with the PANAS-X Joviality subscale, and PLAY itself with the ANSS PLAY subscale. Sociality was measured with co-presence and social presence, and reciprocity with the proportion of apple-related interactions initiated by the participant.

For each measure with three readings, an omnibus test (repeated-measures ANOVA or Friedman test) was run first, and Holm-Bonferroni-corrected pairwise comparisons were made only where the omnibus test was significant. SAM scores at P3 were tested against a threshold at the upper third of the scale. Joviality and ANSS PLAY were each compared with the next-highest subscale, and behavioural indicators were correlated with ANSS PLAY ratings. Interaction frequency was compared across four two-minute blocks of the session with a Friedman test.

Source code

AppleInteraction_source.zip (182 KB): C# scripts, the Unity project settings, and the package manifest.

Each archive holds the C# scripts, the Unity project settings and the package manifest, which is enough to read the code and see which Unity version and packages it was built with. Models, textures, audio and other third-party assets are left out to keep the archive small.