Gallery of Agent Behavior as a Driver for Novelty and Creativity in Human-Robot Co-Drawing

Describing Agent Behavior Via Prompt Design

To describe a range of behaviors, we rely on intentional visual/semantic dissonance/congruence to provoke lateral problem-solving and break creative fixation. Here are example co-drawing outputs from six artists with our nine prompt designs.

Visual

differentneutralsimilar

similarneutraldifferent

Semantic

Co-drawing Outputs from the Physical Robot Agent

With four focused group participants, we asked them to draw an open collaborative landscape drawing over alternating turns. Particpants were given various drawing implements such as color sharpie, highlighters, pencil.

Co-drawing Output with Online Agent

With twenty particpants, we performed preliminary study using browser-based painting app with tablet and stylus.

Conditions:

• Adversarial: Agent introduces high visual and conceptual divergence (visually, semantically different).
• Collaborative: Agent matches user style and theme (visually, semantically similar)

Hover over a drawing (tap on a touch screen) to replay how it was made. Blue strokes: participant's. Red strokes: co-drawing agent's.

Adversarial

Collaborative