Collaborative Research: Improving Worker Safety by Understanding Risk Compensation as a Latent Precursor of At-risk Decisions

员工可能会成为某些认知偏见的牺牲品，将其视为捷径，从而导致判断错误和风险决策，比如风险补偿。风险补偿偏差认为个体通过调整自己的风险行为来达到潜在风险和收益之间的平衡，从而维持一个目标风险水平。风险补偿源自外部（例如，任务或环境相关）和内部（例如，个人特征）来源，最终影响个人（审慎的、情感的和经验的）风险感知，作为健康和安全相关行为和某些风险决策的中心预测因素。在建筑安全环境下，风险下的决策研究主要集中在个体层面。然而，根据社会影响和行为意向理论，同事的冒险行为是工作场所员工冒险行为的“额外动机”。因此，研究建筑环境中的风险补偿效应会变得更加复杂，因为建筑工人是群体工作，同事行为会影响与安全相关的行为。此外，热暴露和随后的热应激的影响可能转化为由身体不适、疲劳和警惕性降低引起的伤害风险增加，从而影响工人的情绪状态和风险感知，并导致认知失败、误解危险和忽视预防行为。因此，这个多学科项目通过整合心理科学、人工智能（AI）和建筑安全方面的进步来解决这些差距，提供了一个新的理论平台和经验过程，以了解干预后工人决策动态的潜在变化，以更好地保护工人免受伤害。本研究的具体目标是：(1)检查个人的特征和心理状态，以及任务和环境因素（例如，时间压力，极端高温）对工人风险决策的影响程度；(2)确定风险补偿偏差对团队风险感知、决策和工作行为的影响；(3)开发一个多维人工智能模型，利用有限的属性，包括个人、任务和环境相关因素，识别有风险的工人并解释他们的风险决策。为了实现这些目标，一个由被动触觉和环境模式组成的多传感器沉浸式360混合现实环境被用来提高工人的存在感，捕捉他们在当前和未来各种施工任务中对安全特征的真实反应。定性和定量相结合的措施有助于调查工人风险补偿行为和决策的潜在机制。测量方法包括位置跟踪传感器、基于视觉的传感器、无线神经心理和认知脑监测（fNIRS）、眼动仪、光体积脉搏描记仪（PPG）和皮肤电反应（GSR）心理生理传感器、半结构化访谈、人口统计和心理调查。收集到的数据构成了使用基于代理的建模模拟的有关工人行为变化的信息，并用于开发多维预测模型，以最大限度地减少风险补偿的可能性，并防止事故和伤害。项目成果有可能影响全国产业的绩效，并为加强国家研究和教育基础设施创造一个新的平台。它们为成千上万的美国工人提供了保护机制，并在美国每年节省了大约数十亿美元的财务成本。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

Workers may fall prey to certain cognitive biases as shortcuts that result in judgment errors and risky decisions, such as risk compensation. The risk-compensation bias argues that individuals adjust their at-risk behaviors to achieve a balance between potential risks and benefits and thereby maintain a target level of risk. Derived from external (e.g., task or environmental-related) and internal (e.g., individual characteristics) sources, risk compensation ultimately influences an individual’s (deliberative, affective, and experiential) risk perception as a central predictor of health and safety-related behaviors and certain risky decisions. Decision making under risk is mainly studied at the individual level in the construction-safety setting. However, drawing on social influence and behavioral intention theories, coworkers’ risk-taking serves as an “extra motive” of risk-taking behavior among workers in the workplace. Thus, studying the risk-compensation effect in the construction environment can become more complicated given that construction workers work in groups, and coworker behavior can influence safety-related behavior. Furthermore, the effects of heat exposure and subsequent heat stress might translate into an increased risk of injury caused by physical discomfort, fatigue, and reduced vigilance that can influence worker emotional state and risk perception, and lead to cognitive failure, misperceiving hazards, and neglecting precautionary behavior. Accordingly, this multidisciplinary project addresses these gaps by integrating psychological science, artificial intelligence (AI), and advances in construction safety to deliver a novel theoretical platform and empirical process to understand the latent changes in worker decision dynamics following an intervention for greater protection from injury.The specific objectives of this study are to (1) examine the extent to which individuals’ characteristics and psychological states, along with task and environmental factors (e.g., time pressure, extreme heat) influence workers’ at-risk decisions; (2) determine the role of risk compensation bias on team risk perception, decision making, and work behavior; and (3) develop a multidimensional AI model to identify at-risk workers and interpret their risky decision-making, using limited attributes including individual, task, and environmental-related factors. To achieve these objectives, a multi-sensor immersive 360 mixed-reality environment that consists of passive haptics and environmental modalities is used to raise the workers’ sense of presence, capture their realistic responses to safety features during various current and future construction tasks. A combination of qualitative and quantitative measures serve to investigate the underlying mechanisms of workers’ risk-compensatory behaviors and decisions. The measures derive from location-tracking sensors, vision-based sensors, wireless neuropsychological and cognitive brain monitoring (fNIRS), eye-tracker, photoplethysmography (PPG) and galvanic skin response (GSR) psychophysiological sensors, semi-structured interviews, demographic, and psychographic surveys. The collected data constitutes information about workers’ behavioral changes simulated using agent-based modeling, and used to develop a multidimensional predictive model to minimize the likelihood of risk compensation and to prevent incidents and injuries. The project outcomes have the potential to impact the performance of a nationwide industry and create a novel platform for enhancing the national research and education infrastructure. They advance protection mechanisms for thousands of American workers and save estimated billions of dollars in financial costs per year in the United States.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.