Food delivery riders are overrepresented in road crashes. Arguably, the increased risk experienced by food delivery riders is linked to the working conditions offered by the “gig economy”. Research is needed to fully understand the safety-related issues this vulnerable group of road users face daily and identify opportunities for counter measures. In this investigation, we proposed a new theoretical model to explain the risky behaviour of food delivery motorcyclists based on the well-established Job Demands-Resources (JD-R) model. Following the JD-R, we considered the impact of job demands (job aspects that require sustained effort) and job resources (job aspects that help achieve work-related goals, reduce job demands and stimulate personal development) on the risky riding behaviours of food delivery motorcyclists. The JD-R model was also extended with three constructs, including personal demands, personal resources, and perceived safety risk to explore the role of individuals' within-person aspects. The developed model was tested using data collected from 554 food delivery riders in the two biggest cities in Vietnam. The results showed that job burnout, job resources, and personal demands directly impact risky riding behaviours, in which job burnout was the most significant predictor. Constructs such as job demands, personal resources, and perceived safety risk were not significant predictors of risky riding behaviours. This research shows that organisation-level factors could be modified to prevent risky riding behaviour. The gig economy industry can do much more to improve the safety of delivery riders.@en

On-demand food delivery involves transport services based on gig-economy models. Food delivery services rely on motorcycles in many jurisdictions, resulting in safety risks. Motorcycles are generally-two-wheeled and therefore inherently unstable. They also lack rider restraint or roll cage to minimise the consequences of a collision. Given the risks of motorcycle food delivery, there is a need to understand how job design may influence safety behaviour on the roads and regulate this economic activity to minimise potential harmful health consequences on the riders. This study investigated the impact of job demands and resources on food delivery riders' compliance with road safety regulations. The job demands-resources (JD–R) model was used as the theoretical framework for this research. Data were collected using a cross-sectional design involving 550 motorcycle delivery riders in two megacities in Vietnam. A structural equation analysis indicated that job demands (e.g., time pressure, work/life imbalance, working environment) and job resources (e.g., social support, feedback) influence, directly and indirectly, job strain, risk-taking attitude, and road safety compliance. Control variables such as age, gender, and income also influenced road safety compliance. This study has critical implications for the food delivery industry that can help achieve sustainable development goals in the global south.@en

Tabular data is one of the most common forms of data in the industry and science. Recent research on synthetic data generation employs auto-regressive generative large language models (LLMs) to create highly realistic tabular data samples. With the increasing use of LLMs, there is a need to govern the data generated by these models, for instance, by watermarking the model output. While the state-of-the-art Soft Red List watermarking framework has shown impressive results on standard language models, it can not be seamlessly applied to models fine-tuned for generating tabular data due to i) column permutation and ii) the task’s nature of generating low entropy sequences. We propose Tabular Red GrEen LiST (T-REST), an adaptation of the Soft Red List watermarking algorithm on tabular LLMs that is agnostic to column permutation and improves detection efficiency by employing a weighted count method that favors columns with higher entropy. Our experiments on 4 real-world datasets demonstrate that T-REST introduces a nonsignificant drop of 3% in the synthetic data quality compared to the non-watermarked data, using the resemblance and downstream machine learning efficiency metrics, while achieving high detection accuracy with AUROC of over 0.98. T-REST is insusceptible to any column or row permutation and is robust against post-editing attacks on categorical columns by maintaining a True Positive Rate (TPR) of over 0.85 when 50% of categorical values are modified.