Researchers have leveraged an unconventional online game, where players herd virtual cattle, to gain insights into human decision-making regarding movement and navigation. This innovative approach may also enable more effective collaboration between humans and artificial intelligence, as well as inform the development of faster-moving robots in the future?
Researchers from Macquarie University in Australia, Scuola Superiore Meridionale and Federico II University of Naples, Bologna University in Italy, and City, University of London in the UK collaborated on a study to explore how dynamical perceptual-motor primitives can inform human decision-making processes.
A Dynamic Programming Model (DPMP) is a theoretical framework that enables us to comprehend the complex processes by which we integrate our actions in real-time, adapting them to the dynamic environment surrounding us. DPMPs have been employed to facilitate insights into our decision-making processes for navigation and task-switching, enabling a deeper understanding of how we navigate between distinct tasks.
In complex settings where diverse individuals converge with stationary and mobile objects, such as a bustling pedestrian area or a sports field, this concept assumes particular significance.
Prior to this discovery, it was commonly thought that our brains rapidly construct intricate cognitive maps of our surroundings and subsequently use these mental representations to plan optimal routes.
Research suggests that rather than creating a detailed plan, we tend to navigate towards our goal, adapting to challenges as they arise along the path.
In the latest research, participants were tasked with undertaking two distinct herding responsibilities: individually relocating either a solitary cow or a group of cows into a designated enclosure.
Researchers monitored the sequence in which gamers herded virtual cows, feeding this data into a decision-making process model (DPMP) to determine if it could replicate the behavior of human gamers.
According to lead writer and PhD candidate Ayman bin Kamruddin, the workforce’s data-driven prediction modeling process (DPMP) successfully replicated the way players moved and even forecasted their decision-making processes with precision.
Researchers examining multi-target decision-making processes found that participants’ initial target choice was typically the one closest in angular distance, subsequent choices were consistently closer to the previous selection, and when presented with a binary option, they tended to choose the cow furthest from the containment zone’s center.
Upon conducting an experiment where we provided the DPMP with these three guidelines for decision-making, it was able to accurately forecast approximately 80% of future choices regarding cow herding, as well as anticipate the behavior of individual members in novel scenarios involving varying numbers of cattle.
Video games that simulate herding behavior are commonly employed in studies such as this because they effectively replicate real-world scenarios where individuals must manage and control multiple entities.
Prior to now, researchers have relied mainly on aerial views of goal animals, prompting concerns that this unnatural perspective may be biasing the results, as it forces participants to make different decisions than they would in real-life scenarios simply due to having a comprehensive overview.
To understand this innovative approach, the team created a novel form of herding sport that simulated a human’s field of view, akin to that experienced in first-person roleplay video games, where players’ perspectives are limited to what they can see through their character’s eyes.
Professor Michael Richardson, a senior writer at Macquarie College’s Efficiency and Experience Analysis Centre, notes that this shift in perspective has profound repercussions.
“Building on previous findings that demonstrate the potential of DPMPs to predict crowd behavior and track moving targets, our study marks the first instance where this model is applied to investigate how humans guide digital characters or robots.”
This incremental approach helps inform the design of even more responsive and innovative solutions.
“Our research underscores the importance of integrating effective collaboration strategies into DPMP frameworks to enable robots and AI systems to better emulate human interactions, social behaviors, and teamwork.”
Additionally, experts suggest that DPMPs could prove valuable in realistic scenarios such as crowd management, evacuation planning, firefighter training in virtual reality, and search and rescue operations, as they can assist in predicting how people will respond and move.
