Want to Experiment on Your Virtual Workers? Ask Them First
Tech companies such as Amazon, Facebook, Netflix and Google are well-known for performing constant experimentation on their digital platforms, with users unknowingly engaging in a whole range of A/B tests while the companies test different tweaks.
But it’s not just consumers who are subject to constant unknowing experimentation. Recent research from Imperial College London highlights how similar experiments are rampant across the various gig economy platforms. The paper examines what impact these experiments might have on the people who depend on gig platforms for their livelihood.
Human Experimentation
The researchers trawled through a couple of decades’ worth of data from a huge digital labor platform and found three distinct phases of experimentation. Each of these phases was characterized by a unique approach to experimentation, both in terms of its nature and the way in which workers responded.
Related Article: Experimentation is the Future of Enterprise Learning
Experimentation began as an explicit, opt-in process that was only performed on workers who had knowingly consented, but it quickly became a more covert process, in which workers were experimented on without their knowledge or consent. This evolved to the point where experimentation was running rampant, with a wide range of experiments underway at any one point.
While experimentation at first increased worker autonomy, this shift then diminished instead — and the final phase normalized that diminishment.
A Key Process
By tracking the evolution of the platform over a long timeframe, the researchers began to understand how important to the company experimentation on workers became.
Drawing from corporate announcements and online forum exchanges, the researchers discovered the platform initiated its initial forays into worker trials around 2007, inaugurating what the authors dubbed the "explicit experimentation regime."
During this phase, the platform publicly shared information about the features it was testing, such as modifications to the "job search" function that improved how clients and workers connected. In an open forum post, the company invited workers to take part in beta testing the new feature and provide their feedback.
Workers were also given the opportunity to suggest areas of the platform to be subject to experimentation and could choose to opt out of any experiment. This phase not only empowered workers with greater autonomy but also set a precedent for what they could expect from the company in the future.
Shifting Habits
A noticeable shift occurred in 2014, when workers on the platform discovered the company was running experiments with the design and display of various metrics in worker profiles without seeking any opt-in approval. While the initial changes seemed somewhat benign, they were the thin end of a wedge that encouraged the platform to grow more ambitious with its use of experimentation.
Learning Opportunities
The constant experimentation on workers later became an official company policy, but the disclosure didn't prompt them to scale back their efforts. What's more, the experiments began to take on a more sinister tone, as some users received automated messages saying their accounts would be suspended if they attempted to communicate with clients outside of the platform.
The researchers described this period as the "concealed experimentation regime," in which experiments were commonly performed without the consent of workers. Workers’ sense of autonomy fell during this phase, as they could no longer control whether they participated. Their choice — their ability to make a choice — ceased to matter.
Running wild
The “concealed experimentation” phase seemed to embolden the company, which then began to significantly increase the number of experiments it was running. The next wave of experiments both covered a wider range of features on the platform and also had no clearly defined endpoint. Indeed, workers would often be subject to multiple experiments at any one point in time.
This process not only diminished the sense of autonomy that workers felt, but also often inhibited their ability to do their jobs. The shift in the company’s attitude was typified by the fact that it offered no acknowledgment that experimentation was taking place, much less asking for permission to do so.
Related Article: Give Employees a Say in Future Technology Decisions
Interestingly, the shift was met with passivity by the workers themselves, who didn't either voice concern publicly or simply began to vote with their feet and leave the platform. The researchers believe the two sets of actions — either saying nothing or bailing on the platform — reflected a grudging acceptance from workers that they would be experimented on consistently throughout their time on the platform.
That sense of resignation should not be taken to mean this is good practice, however. I've written previously about the importance of establishing trust with employees when monitoring what they do remotely, and this study is a perfect example of a platform abusing the trust it had initially established with workers.
As the study shows, when workers cotton on to the fact that monitoring and experimentation are not designed in a way to be either inclusive or mutually beneficial then morale and productivity will inevitably fall — and turnover will inevitably rise.Learn how you can join our contributor community.
About the Author
I currently advise the European Institute of Innovation & Technology, am a researcher on the future of work for the University of East Anglia, and was a futurist for the sustainability innovation group Katerva, as well as mentoring startups through Startup Bootcamp. I have a weekly column on the future of work for Forbes, and my writing has appeared on the BBC and the Huffington Post, as well as for companies such as HCL, Salesforce, Adobe, Amazon and Alcatel-Lucent.
Connect with Adi Gaskell:
Featured Research
Buyer's Guide
Read Now
Research Report
Read Now
eBook
Read Now
Research Report
Read Now
White Paper
Read Now
Research Report
Read Now
Guide
Read Now
Guide
Read Now
Related Stories
Entries for the 2023 Reworked IMPACT Awards are now Open
