How Mobile Crowdsourcing Can Improve Occupational Safety

With 150 workers dying each day from hazardous working conditions, work safety continues to be a serious problem in the U.S. Using mobile technology to collect information about workplace safety conditions from those on the ground could help prevent serious injuries and save lives by accelerating the ability to spot unsafe conditions. The convergence of wireless devices, low-cost sensors, big data, and crowdsourcing can transform the way we assess risk in our workplaces.Government agencies, labor unions, workers’ rights organizations, contractors and crowdsourcing technology providers should work together to create new tools and frameworks in a way that can improve safety and provide value to all stakeholders.

Crowdsourcing (the act of soliciting help from a distributed audience) can provide a real-time source of data to complement data collected by government agencies as part of the regulatory processes of monitoring workplace safety. Having access to this data could help government agencies to more effectively monitor safety-related legal compliance, help building owners, construction companies and procurement entities to more easily identify “responsible contractors and subcontractors,” and aid workers and unions in making more informed choices and becoming better advocates for their own protection. Just as the FitBit and Nike Wristband provide individuals with a real-time reflection of their habits designed to create the incentive for healthier living, crowdsourcing safety data has the potential to provide employers and employees alike with a more accurate picture of conditions and accelerate the time needed to take action.

We Don’t Know Who Is at Risk and Where

By the time you finish reading this, one person will likely have died from dangerous working conditions somewhere in America. The extent of the problem is alarming:

• In 2014, 4,821 working people were killed on the job — a rise of 2% versus the previous year — and approximately 50,000 died from occupational diseases.
• The true toll of hazardous workplaces is likely to be much greater than what is being reported. “OSHA (Occupational Safety and Health Administration) believes that many severe injuries — perhaps 50% or more — are not being reported.”

Why haven’t we been able to make significant progress towards making our workplaces safer?

For one, federal enforcement agencies simply do not, and will never, have enough resources to inspect every site. Secondly, penalties for workplace safety violations are still far from being detrimental for many companies to take matters more seriously. In construction, a notoriously dangerous industry, incentives are often given for early completion of projects, whereas failing to meet project deadlines can result in penalties. Budgetary pressure, especially for publicly-funded construction, encourages the use of low-bid contracting, often resulting in substandard work and loss of life and limb. Furthermore, the changing nature of the workplace, with more workers being employed through subcontractors adds to the enforcement problem by making the chain of control and management harder to trace.

Most importantly, however, we have no real-time information about the safety conditions on worksites. Under a new rule that is expected to go in effect in July 2017, OSHA will expand the reporting obligations for certain high-risk companies by requiring electronic reporting of many types of work-related injuries and illnesses within 24 hours. However, many small and mid-size employers will be exempt from the rule, while underreporting will most likely continue to be an issue.

Previous studies on indicators of workplace safety have found that while lagging measurements can provide data about incidents after the fact, the value of these metrics as future predictors for safety is questionable. As such, it is highly likely that the available regulatory data will continue to be insufficient to predict current and future safety risks of workplaces. In order to improve workplace safety (and bring better transparency into identifying who is a responsible employer and who isn’t) there is a need to document who is at risk and where, before rather than after the fact.

Enter the Wisdom of Crowds

Tapping into the “wisdom of crowds,” where the “crowds” are the workers themselves who are on site can be an efficient way to fill this critical knowledge gap. Crowdsourcing can provide a more robust understanding of the worksite safety risk associated with a particular company by augmenting readily available historical data which by itself is inadequate to measure risk.

Such knowledge discovery approaches are already being used in other domains. Ushahidi, for instance, asks the crowd to find and report instances of urban disrepair or ethnic violence, which governments then use to allocate city resources or deploy peacekeepers. Another example is PulsePoint, an app that alerts nearby CPR-trained first responders to come to the rescue for faster emergency response.

Similar approaches are starting to emerge in the field of occupational safety. One example is SeeClickFix, where the crowd is asked to report instances of urban disrepair which governments can then use to allocate city resources. In collaboration with Workers Lab, the company recently developed Worker Report app to test “whether workers across an array of industries would use technology to report wage theft and health and safety violations to local workers’ rights groups”.

Mobile, and more specifically SMS in particular, is key in capturing audiences in the right context and a growing population of workers now have smart phones. Even though many worksites incorporate a no-mobile policy while at work, it is still the right medium to engage workers when they notice a threat to their safety. Mobile would allow for easy, instant reporting at the worker’s convenience through a communication medium that s/he already uses; therefore removing potential entry barriers.

Here is how such a mobile-based early warning system could work:

In order to submit a safety report, a worker would send a text to a number (or messenger account for those with a smartphone), which would trigger a series of simple, yes/no type safety questions. The questions would evaluate not only safety behavior and attitudes, but also safety conditions to be able to form an accurate assessment of the situation.

A website would show safety profiles of companies, including an aggregated safety score and the historical trend, based on reports collected for that company’s worksites. The real-time reports could be combined with government-collected data to offer a more robust picture.The company page would reveal key facts about the company, a company safety score that would average that company’s all site safety score(s), as well as the safety trend over time.

Users would be able to subscribe to alerts to receive notice when a new report is posted for that company. This way, contractors would be able to track their companies and have a chance to investigate when a safety report has been submitted.

Designing for Crowdsourcing Work Safety Conditions

Done right, crowdsourcing an early-warning workplace safety system is beneficial to all key stakeholders in the system:

• For workers, it provides a way to report on both safe and unsafe conditions and increase the likelihood that action will be taken sooner when something is unsafe.
• Unions and advocates can find out which companies and contractors pose a threat to worker safety before an accident happens.
• Employers can learn how they compare to competition on safety and health and discover opportunities to improve safety and health, satisfaction, and productivity across their operations.
• For responsible employers, the positive scores associated with their brand can provide a competitive advantage in the search for contracting and subcontracting opportunities.
• Crowdsourced data can enable efficiency for federal and state enforcement agencies, guiding them to allocate inspectors in a smarter, targeted way and potentially transforming enforcement.

Given that many are reaping the benefits of crowdsourcing for various purposes today, why has there not been an effort to introduce crowdsourcing in the workplace to enhance safety? There is a perceived, albeit inaccurate, tension — one played up by those who have an incentive to drive costs down — between investing in safety and creating more jobs. There is also a valid concern that creating such a whistleblowing mechanism will unfairly harm good companies by providing too easy a way for disgruntled workers and others to complain. Ultimately, however, for a system like this to be effective it has to meet the needs of all key stakeholders. Human-centered design, namely using ethnographic research to understand the needs of workers, safety experts, workers’ comp insurers, unions, and occupational health and safety agencies, is key to designing a service that works in practice and will be used.

Through a series of interviews, we have developed a set of principles which should inform the design of a mobile crowdsourcing prototype and a website to publish the data:

1. GO MOBILE: Among U.S. adult population, 92% use mobile phones, while 68% use smart phones. However, this doesn’t necessarily need to suggest apps, particularly when the reach is important. Convincing someone to download and use an app is difficult and costly. Instead, there is a need to test communication channels people already use, such as SMS and messaging apps

2. WIN-WIN-WIN: Challenging as it is, in order to maximize its impact, such a safety warning system would need to provide value to all key stakeholders in the system, including employers, by ensuring that crowdsourced information results in valuable outcomes rather than merely finger-pointing.

3. EMPLOYEES AS THE CROWD: Workers are not only the population most vulnerable to shortcomings in workplace safety practices, but are also in the best position to share information that their employers might be hesitant to share.

4. TEST THE INCENTIVES: One of the most critical questions imperative to the success of a crowdsourcing initiative is how to create and sustain the engagement of the crowd. Our initial qualitative field studies confirm what we know from the research literature (Huttenlocher, Brabham, Kittur et al.), namely that intrinsic incentives are more valuable than extrinsic ones to motivate participation. At the same time, running a pilot is instrumental in being able to empirically test the right prime(s) through A/B testing.

5. PRIVACY BY DESIGN: Our focus groups with workers revealed that one of the biggest barriers to safety reporting is the fear of retaliation by the employer. To ensure engagement, it is absolutely critical to protect worker privacy (i.e., through data anonymization).

6. POSITIVE FRAMING: While it is obviously important to identify which companies are at risk, it is equally, if not more, important to celebrate and prominently feature responsible employers in order to encourage positive competition.

7. PREVENT EXPLOITATION: In order to earn the buy-in and engagement from employers, three key factors must be reflected in the design: the presence of clearly communicated safeguards to prevent gaming the system; the ability to respond to the safety reports and take necessary action (closing the feedback loop); and the featuring and highlighting of responsible companies.

8. STRONG INSTITUTIONAL SUPPORT: Although a bottom-up information collection practice, a crowdsourcing process requires a strong institutional owner whose reputation can help to motivate participation. For instance, while they are most at risk, non-unionized workers are also the more difficult population to reach because of the lack of contact information, thus necessitating a heavy reliance on unionized workers. Therefore, the support of labor unions is integral for a successful scaling of a crowdsourcing mechanism that aims to engage unionized workers.

9. ASK THE RIGHT QUESTIONS: Since the questions would be asked via SMS and would go out to workers from different backgrounds, we established the following principles to guide the creation of the questionnaire:

• No more than 10 questions, to maximize engagement;
• Simple wording that prevents any misunderstanding;
• YES/NO type questions to make it easy to respond and encourage completion;
• Mix of questions to assess both safety behavior and perceived conditions.
• A/B testing to ensure best length and wording that maximizes response rate and quality

10. TEST AND ITERATE: Designing such a crowdsourcing framework has taught us that the challenge is not only determining the design and technology to be used, but also engaging all stakeholders. Running multiple pilot experiments with organizations is necessary to fine-tune the interaction and language that is used, as well as to demonstrate the value and impact of such a system to different stakeholders.

Challenges

Maximizing reach and engagement

Reaching a critical mass will be important to generate a steady flow of data for a meaningful number of sites. This requires effort from worker organizations and/or regulatory agencies to promote the service and build awareness. In this context, reaching the most vulnerable population, non-union workforce, is a difficult task.

Another challenge with maximizing engagement has to do with closing the feedback loop. Who does what with the data and who would be responsible for acting on the reports is vitally important in the context of workplace safety. Someone must take action on an incoming report and provide feedback to the initiator. Nothing will kill engagement faster than a report that does nothing other than generate data for some database.

Preventing Exploitation

Another challenge is the need to have enough reports for a site to be able to validate the accuracy of the information, while having design mechanisms in place to prevent gaming the system. Contractors will need reassurance that they’re not being falsely accused by a disgruntled worker or even by a competitor who is trying to harm their reputation, both valid and understandable concerns. All this needs to be carefully taken into account while designing the system.

Engaging Stakeholders

A final key challenge is the engagement of all necessary actors. Engaging public institutions in experimenting and embracing crowd technology tools and new frameworks is no easy task. Our experience working with labor unions has revealed that some unions are hesitant to experiment with new tools which they perceive to be a potential threat to their contractor relationships. Hence, the development of a successful crowdsourcing model would require close collaboration with unions to address such concerns and secure their engagement.

Looking Ahead

We believe that mobile crowdsourcing is the future of workplace safety and can imagine a future where every job site, not just construction, has a designated phone number that feeds a website where data about that workplace can be visualized and compared to that of other, comparable workplaces.

Enlisting the aid of workers in sharing site-specific safety information can augment readily available safety violation data and paint the most accurate picture of workplace safety. Through easy, safe, and anonymous reporting of safety conditions, we can empower workers to speak up without fear, while providing unions, government safety and health agencies, workers’ compensation insurers, and others with vital information on which contractors pose the greatest risks and where they are currently working. Public visibility of safety conditions across sites can be a stronger motivation than federal penalties for companies to improve safety.

There is no reason why in the digital age we should be regulating companies based on year-old data when we can create real-time feedback mechanisms. It will not be easy to overcome the concerns people will have about whether reporting safety conditions might negatively impact their job security. With real time data, however, companies will have the raw material they need to spot and stop problems before they start, dramatically lowering their costs and staving off invasive inspections from regulators.

In order for this to happen certain conditions will need to be met:

First, the stakeholders — regulatory agencies, worker rights organizations and contractor associations — need to work together to align on the objectives.

Second, they need to work with design partners to pilot and test different methodologies and platforms to develop a practical approach.

Finally, unions and worker centers have a critical role to play in creating awareness about the opportunity and encouraging their workers to (anonymously) share information. Even though the most vulnerable populations are non-union, the large worker network and contractor relationships of unions put them in a unique position to test and learn what works best.

By asking workers to share the missing information on current safety conditions of worksites, and through connecting this information to the respective companies, mobile crowdsourcing has the potential to enhance corporate transparency, reduce risk and help save lives.

co-authored by the GovLab’s Batu Sayici & Beth Simone Noveck