The investigation of approach-avoidance tendencies has traditionally relied on computer-based technologies that primarily characterise human behaviour through reaction times. However, these technologies are unable to accurately quantify other kinematic variables such as hand speed and movement direction. To address these limitations, novel robotic devices have been developed, providing more diverse and accurate quantitative assessment of human behaviour. This technical report presents an adaptation of the approach-avoidance task on the Kinarm, a robotic platform designed to track upper limb movements as participants interact with a virtual environment. This variant of the approach-avoidance task assesses the movement of both arms in twelve directions of reach. In addition, resistive loads can be applied to investigate the role of physical effort in approach-avoidance tendencies or to support rehabilitation protocols. Data and analyses from a pilot sample (n = 5) highlights the capabilities of the Kinarm Approach-Avoidance Task (KAAT).