Beyond Crowdsourcing: Flash Organizations

Crowdsourcing has been the flavor of the month for several years now, and outsourcing has become a way of life for many workers everywhere. But there is a huge gap between the trivial microtasks typical of early crowdsourcing and most work. Even as digital gigs have become standard for many “creatives”, there are still many tasks that seem to require more than a mash up of freelancers.

Not to fear, we can make an app for that, too.

Melissa Valentine and colleagues as Stanford published an award-winning paper this month, describing “Flash Organizations: Crowdsourcing Complex Work by Structuring Crowds As Organizations,” [2]

The basic idea is to replace the completely flat social structure of simple crowdsourcing with a more complex structure that resembles conventional organization structures. This is implemented in a platform they call “Foundry”.

Something like Mechanical Turk or Task Rabbit has basically two roles, director (who specifies the work and pay rates) and worker (who performs tasks for pay). There may be wrinkles, such as a third party for quality control, but it’s a pretty simple graph, and the ‘roles’ are very generic.

The Stanford group elaborates this with specialized roles, plus teams and hierarchies. These concepts are familiar from organizations, and have actually been used in digital support for work for many, many, many years.

One contribution is to support more complicated descriptions of work roles. The idea is to be able to abstractly create a complicated organization including specialized experts and teams, which can be rapidly instantiated by recruiting workers with the requisite skills. The advantage is to enable workers who have never met to rapidly “to coordinate using their knowledge of the roles rather than their knowledge of each other.” ([2], p. 3523)  (This isn’t exactly original:  in fact they use Upwork, which already implements these roles, along with simplified hiring.)

The second contribution is to support rapid reconfiguration of the organization as work progresses. This is done by using techniques familiar from version control systems. The (digital model of the) organization can be branched and merged, to maintain a centralized coordinated consensus in the face of multiple parties making changes. (The ‘diffs’ are computed on a graph of objects, not on text or code.)

That’s all great, but how do you fill the slots with human workers?

Apparently, this is tackled by a virtual temp service, which maintains “panels” of “pre-vetted” workers. The current implementation uses Upwork’s process to create these panels.

When the organization is booted up, and as it changes, the platform automatically pulls workers out of the relevant panels, as needed. Since the screening has already been done (outsourced), “hiring” takes no time at all. Poof! (This turns out to be one of their key metrics for the platform.)

The paper describes three experiments using the platform to create virtual organization and execute software projects. The descriptions of the projects sound a lot like how these kind of projects would happen within a large organization. That is, when the group decides it needs a new expert or team for a part of the project, relevant workers are called in to tackle the task. In this sense, the platform is successfully delivering the advantage of being part of IBM or Google, in a crowdsourcing system.

“Turning freeelancers back into corporate drones” is probably not a tag line they hope for. 🙂

This platform makes it really easy for managers to fiddle with the project tasks and organization. Anyone who has dealt with pointy-haired managers recognizes the risk here. Indeed, the paper hints that some workers thought management needed to be more thoughtful about changing the organization and tasks.

“Making it easier to mismanage whole organizations” is another motto this project would not hope for. 🙂

The authors recognize that this model has limitations, and they discuss a few.

Naturally, the lack of organizational context and loyalty may be an issue for many cases. I’ll add that intellectual property and security concerns would preclude the completely open hiring process in many cases.  For example, the paper cites a time when a manager recognized a need for expertise, so he “hired a web security engineer in Egypt to train himself” on the relevant compliance process ([2], p. 3530).  I leave it as an exercise for the reader to list all the problems with this move.

This platform is obviously geared toward work that is organized as repeated gigs. Their models are film production and disaster response. These industries feature permanent organizations that perform the same complex task over and over. This enables the development of a well understood set of roles, and pools of qualified workers, and a need to boot up, execute, and tear down project specific organizations.

It’s not clear that other types of work can be shoe-horned into this model. For example, formal education might be thought of this way. But many would consider the task to extend over many years, and want it to adapt to the needs and preferences of the students (not the school managers). Hiring teachers in minutes isn’t nearly as important as maintaining a sustained progress over the student’s career.  For that matter, I don;t think that personal contact with role models, mentors, and peers can’t be outsourced to anonymous experts on the internet.

This platform requires this sort of well-understood expertise, such as “Android App Developer”. We can train and certify people for these skills. In recent years, we have even been able to retrain workers as technology evolves.

I’m not expecting there to be pools of interchangeable experts in every needed skill.  There are plenty of skills that are harder to train, and harder to certify. Skill at writing. Math chops. Customer service. Musical creativity.

Finally, one could consider the philosophical question of how you treat people. This model explicitly treats workers as interchangeable units. The Foundry created organization is not a group of people, it is a digital model that can be fiddled with like a video game–except the ‘NPC’s are actual humans. Workers do not know each other, and have pretty much no personal stake in the outcome. Bosses neither know nor care about the workers.

One of the benefits of working in an organization face-to-face, is that the workers do know each other and, ideally, care for each other. If you have worked in a big organization, you have probably done things to help another worker or the whole organization—regardless of your formal role. Building and sustaining a good organization can be the most rewarding part of work, and can be motivating far beyond paychecks.

On this point, it is instructive that one of the acknowledged models for the Foundry is film production.  Hollywood is notoriously a terrible place to work, especially for the skilled workers who are treated as interchangeable units. The top honchos and big stars may have fun (and make a fortune), but everybody else is abused and bullied and many are just hanging on from gig to gig.

Is this the model we want to follow as the future of work?

  1. Taylor Kubota, Stanford researchers develop crowdsourcing software to convene rapid, on-demand ‘flash organizations’, in Stanfor – News. 2017.
  2. Melissa A. Valentine, Daniela Retelny, Alexandra To, Negar Rahmati, Tulsee Doshi, and Michael S. Bernstein, Flash Organizations: Crowdsourcing Complex Work by Structuring Crowds As Organizations, in Proceedings of the 2017 CHI Conference on Human Factors in Computing Systems. 2017, ACM: Denver, Colorado, USA. p. 3523-3537.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s