Scientists Hack Pac-Man to Make Physical Therapy Less of a Chore

Table of Links

• Abstract and 1 Introduction
• 2 Related Work and 2.1 Modifying Games for Rehabilitation
  • 2.2 Input control performance measure and modelling and 2.3 Summary
• 3 Motivation from a rehabilitation context and 3.1 Observations and Brainstorming workshops
  • 3.2 System description
  • 3.3 Interaction parameterisation
  • 3.4 Summary
• 4 Experimental Study
• 5 Modelling User Behaviour and 5.1 Behavioural Features
  • 5.2 Gameplay Reference Model
  • 5.3 Comparing NLL to NSCORE
  • 5.4 Effect of design parameters on NLL
  • 5.5 Sampling Period
• 6 Discussion and 6.1 Design workshop and tools for modelling
  • 6.2 Interaction parameters adaptation
  • 6.3 Modelling Gameplay
  • 6.4 Limitations and Future work
• 7 Conclusion
  • 8 Acknowledgements and References

ABSTRACT

Modified digital games manage to drive motivation in repetitive exercises needed for motor rehabilitation, however designing modifications that satisfy both rehabilitation and engagement goals is challenging. We present a method wherein a statistical model of baseline gameplay identifies design configurations that emulate behaviours compatible with unmodified play. We illustrate this approach through a case study involving upper limb rehabilitation with a custom controller for a PacMan game. A participatory design workshop with occupational therapists defined two interaction parameters for gameplay and rehabilitation adjustments. The parameters’ effect on the interaction was measured experimentally with 12 participants. We show that a low-latency model, using both user input behaviour and internal game state, identifies values for interaction parameters that reproduce baseline gameplay under degraded control. We discuss how this method can be applied to systematically balance gamification problems involving trade-offs between physical requirements and subjectively engaging experiences.

Keywords Game; Rehabilitation; Interaction Design; User modelling.

1 Introduction

The engaging nature of interactive systems, especially games, has the potential to improve rehabilitation sessions, as long as their interaction design aligns with patients’ needs and capacities. One has to find and assess the relevant interaction design parameters that could be adapted to fit the rehabilitation requirements and specifications (Lopes and Bidarra, 2011).

Gamification (Raczkowski, 2014) seeks to incorporate “video game elements in non-gaming systems to improve user experience and user engagement” (Deterding et al., 2011). The use of games has been shown beneficial for patients in a wide variety of rehabilitation contexts. Focusing on the upper limb, Sietsema et al. (Sietsema et al., 1993) showed that an interaction with a Simon game produced significantly more range of motion than a standard rote exercise for patients with traumatic brain injury. However, when games are not initially intended for training purposes, the interaction between the game and the player must often be altered to meet rehabilitation goals. For example, Half-life 2 (Valve Corporation, 2004) was partially controlled through a recumbent bicycle, by linking the pedalling speed to the move forward action of the playable character (Ketcheson et al., 2016), for purposes of player exertion. Then additional parameterisation of the interaction is typically needed to accommodate specific patients ability (Barrett et al., 2016). For instance, content from the game of Fruit-Ninja (Halfbrick Studios, 2010) was removed to cater for patients’ reduced mobility during arm rehabilitation (Khademi et al., 2014). These modifications are often introduced using heuristics (Pirovano et al., 2016; Munoz et al., 2018) or presets (Ketcheson et al., 2016; Chatta et al., 2015) but ˜ the extent to which they impact gameplay is difficult to predict.

In this paper, we propose to develop models of baseline gameplay, captured through game sessions from a control player group, to serve as a reference against which interaction modifications can be evaluated, alleviating the reliance on heuristics or presets.

Our contributions are twofold. First, we design a system for upper limb rehabilitation based on minimal game alterations and parametric interaction design. Second we validate a novel metric for predicting game difficulty using the designed system, based on models of baseline gameplay. These contributions benefit to the field of game design for rehabilitation by providing a new method for balancing game alterations compatible with a computational approach.

This research is motivated by field work with occupational therapists leading to the development of a system for movement-based rehabilitation using the game of Pac-Man. A series of workshop and testing sessions grounded our design choices: interaction adaptation is performed with the modifications to the game’s time rate and the optimisation of a newly designed input device physical properties. In addition, our contributions benefit to the field of Computational Interaction in HCI (Oulasvirta et al.), which relies on models to gain insight into interactive systems. We propose a model of baseline gameplay, constructed from low-level variables, that correlates with in-game score. We show that common measures of performance, such as in-game score, are prone to high latency and inter-user variability, and propose instead learning a statistical model of normative control behaviour.

The paper is structured as follows. In the next section we present previous work in modifying games for rehabilitation and related research to control input adaptation and user performance modelling. We then report the work carried out from an participatory workshops with occupational therapists and the resulting designed systems. An experimental study involving 12 participants follows wherein we investigate how a new input modality affects the user experience and users’ motions. Finally, to develop of a low-latency proxy for interaction difficulty, we detail a statistical modelling of collected data which is shown to correlate with game score.