News Release

Computing the ways of learning by doing

Grant and Award Announcement

Singapore Managment University

SMU Associate Professor Cheng Shih-Fen

image: New research, led by SMU Associate Professor Shih-Fen Cheng, is utilising big data from Singapore's transport gig economy to evaluate how and at what rate individuals learn over time. view more 

Credit: Singapore Management University

By Alistair Jones

SMU Office of Research & Tech Transfer– How can we improve the way we do our jobs? A popular strategy is to sign up for online courses, or attend special seminars and workshops.

But according to many education researchers and experts, it is learning by doing (LBD) that has proven to be more effective than any other kind of training. LBD is hands-on experiential learning, also known as learning in the flow of work.

It is not a new idea. LBD gained prominence in progressive pedagogy around the turn of the 20th century through the theories of John Dewey, an American philosopher, psychologist and educational reformer. But as far back as 350 BCE, Aristotle had written, “for the things we have to learn before we can do them, we learn by doing them”.

LBD works, but while it is conceptually simple and intuitive, its sources and drivers remain a mystery. As a result, even when a firm intends to facilitate LBD among its employees, it is not clear how to effectively achieve it.

Now, a new research project led by Shih-Fen Cheng, an Associate Professor of Computer Science at Singapore Management University (SMU), seeks to bring clarity to the mechanisms of LBD and reveal what it is that enables LBD to happen for an individual, and how it can be promoted.

The study, using the computational tools of big data to examine LBD in the operation of Singapore's transport gig economy, was awarded a 2020 Ministry of Education (MOE) Social Science Research Thematic Grant.

Potential issues

So, why have the sources and drivers of LBD remained a mystery?

“Historically, LBD effects have been measured as the rise in productivity in response to cumulative time spent working,” Professor Cheng says.

“There are two potential issues with this commonly accepted definition. First, what [is of interest] in the study of LBD is a worker’s skill, which is unfortunately unobservable in most cases. In its place, most researchers choose to measure productivity instead, which could be seen as a good proxy for the underlying worker skills for many simple job functions.

“However, this definition is far from adequate, particularly when jobs are complex and the relationship between skills and productivity is either noisy or non-linear.

“Second, though one potential way to identify productivity-based LBD is to look for historical events where investments (such as labour, capital and technology) are stalled over long periods – which enables the testing of the impact of LBD on productivity – we do not usually observe such a case.

“Furthermore, even in historical cases where such a lack of investment occurred, changes in unobserved variables such as product quality, technology or organisation may have more of an influence on increasing productivity, thereby diluting the contribution of LBD. Consequently, the analysis to evaluate how and at what rate individuals learn over time has been difficult,” he says.

Towards a solution

Professor Cheng and his team propose a novel big data framework to measure the LBD effects for workers in the transport gig economy in Singapore by mining drivers' microscopic movement traces and trip fulfilment. The high-fidelity data comes from the Land Transport Authority (for taxi drivers) and Gojek (for ride-hailing drivers).

“We plan to computationally quantify not just the evolution of individual drivers’ productivity, but also their skills in anticipating demands and competition from other drivers, measured using models deriving from the area of behavioural game theory,” Professor Cheng says.

The basic idea of the model is the introduction of the 'level of reasoning' – representing skill – which refers to the number of iterations an agent can perform in strategic anticipation of other agents. The detailed data enables this to be measured at an individual level over time, rather than as just a snapshot.

The researchers have identified four main channels through which drivers can increase their earnings: labour supply (hours worked); street hail trips (only for taxis); booking trips (for both taxis and ride-hailing drivers); and strategic labour supply. The study will extract contributions from the different channels on reasoning levels.

The strategic labour supply is of particular interest, since this reflects a driver’s ability to choose the best area to roam in order to increase his/her chances of getting jobs.

“Our conjecture is that the more strategically sophisticated a driver is, the more likely a driver is able to have good supply estimation, and thus be better at choosing the 'right' area to serve,” Professor Cheng says.

And as the collected data covers both the pre-COVID and various phases of COVID close down and re-opening, the researchers hope to discover whether the ability to learn in 'peace time' (drivers shown to have a high level of LBD effects before 2020) could translate into adaptivity in the face of extreme demand shocks.

“[We surmise that] drivers with a higher level of strategic reasoning should be more adaptive, leading to better performance in the face of drastic demand shock. COVID-induced demand pattern changes will be one such representative event in validating this conjecture,” Professor Cheng says.

Technology-based interventions

In earlier work, Professor Cheng was the co-designer of an optimisation-based Driver Guidance System (DGS) for taxi drivers that enabled them to reduce their vacant roaming time by 34 percent during a two-year Singapore trial.

“The use of DGS is a technology-based intervention, which is essentially an algorithm that assists drivers in making better decisions,” Professor Cheng says.

“We are interested in exploring whether assistive technologies can lead to lasting behavioural changes – that is, can drivers actually 'learn' from technologies similar to the DGS,” Professor Cheng says.

An end goal for the researchers would be to try deploying an app-based intervention system to help drivers learn more quickly and improve productivity – that is, to facilitate LBD.

“Our ambition is to use LBD as a theoretic construct to understand how new technologies or platforms can induce different worker behaviours, particularly in the transport/logistic gig economy sector. This covers ride-hailing services, food delivery services, and logistics services,” Professor Cheng says.

“As more workers are moving into these areas, it is crucial to understand how gig workers can be made more efficient through the use of technologies.”

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of news releases posted to EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.