Performance Is All You Need
Outlining the "transformer paper" for Open Work
In 2017, a team of Google researchers kickstarted a revolution with one of the most consequential research papers in (at least) the 21st century.
The transformer paper, Attention Is All You Need, spotlighted how a mechanism for neural networks that was previously just a supplement to a slower, less reliable, and less scalable approach (the “T” in ChatGPT) could replace the old architecture entirely, and uncovering that known but under-utilized principle birthed what we’ve come to know as modern AI.
Today, I’m assembling the team that will write the “transformer paper” for Open Work.
I have a hypothesis that a known but under-utilized principle - portable performance - could become the new core infrastructure of the labor market that unlocks unprecedented value for employees, employers, and society.
But before Open Work has its 2022 “ChatGPT moment” with the public, it needs to have its 2017 “transformer paper” moment with economists, psychologists, and other researchers who can help pressure test the science and turn the thesis into impact.
Here’s are two hypotheses for which this research would need to provide compelling evidence in order to convince the field that this could be the future of work:
H1: Employees with Open Work records experience better hiring outcomes
I’d start with correspondence studies that track outcomes like callback rates, interview invitation rates, and recruiter response rates since they’re well established among labor economists (thanks largely to the canonical study “Are Emily and Greg More Employable Than Lakisha and Jamal? A Field Experiment on Labor Market Discrimination” by Marianne Bertrand and Sendhil Mullainathan) and relatively inexpensive.
The simplest version would be comparing two identical applicants where one of them provides verified work performance and the other does not. The next level could study not just whether the presence of any data makes a difference, but at what levels of performance does it really move the needle. At what point does performance return better outcomes for applicants than education or years of experience? What happens to applicants who are historically challenged in the job market when they add verified work performance (ex. race, ethnicity, gender, nationality, immigration status, criminal history, veterans)?
H2: Employers with Open Work reporting experience better workforce outcomes
If correspondence studies suggest that employers value verified work performance (maybe even as much or more than credentials), employers who are able to hire the high-performers should expect higher productivity, and that certainly could be tested by tracking performance data over time across employers.
However, by definition, not all employers can hire every previous high-performer. If all Open Work does is better identify the best applicants for the best firms to recruit and select, it will likely just increase the gap between well-resourced and under-resourced employers. I’m not sure that’s a net negative impact as many people who might not otherwise get those opportunities would benefit, but this is about more than just a screening tool.
The real power is if we see a rise at every level across the performance spectrum.
Once Open Work is introduced, do we see employees working harder for their current employer, presumably because they know their performance will impact their future job prospects?
Credit scores, 5-star ratings, customer reviews, and auto insurance underwriting have already proved that people will behave differently when they know that the decisions they make today will inform the choices they have tomorrow.
Psychologists have plenty of methodologies in their tool box to test this type of behavioral impact, including real-effort experiments where participants can complete tasks in-person or online where some do so in an Open Work-like environment where their performance is portable versus where performance just disappears.
Real-world pilots of Open Work
In my original essay, The Case for Open Work, I introduce three potential domains for Open Work: prisons, schools, and robotics.
Here are some more groups that operate with high natural turnover that I would approach about piloting Open Work if the aforementioned studies boost the conviction for this model:
Nonprofit enterprises who temporarily hire formerly incarcerated individuals as they transition back to society
Retail and logistics during holiday season
Tourism, hospitality, food, events, and transportation during travel season
Local governments and organizations that run youth job programs during the summer
Work study on college campuses
Staffing agencies, gig platforms, and trade unions that funnel work to their members on a per job/task basis
Success with temporary workforces could unlock more pilots with larger permanent workforces, starting with employers for whom career mobility within and outside their company is a core part of their brand. The macro forces of AI, robotics, layoffs, slowed hiring, and job disruption will likely turn mobility opportunities into table stakes alongside wages and benefits for employers to compete for talent.
If these studies and pilots are successful and there’s growing demand for Open Work from employees and employers, HR tech platforms will have more confidence that investing in, offering, and scaling Open Work will boost their own usage, retention, and new customer acquisition metrics.
Please reach out if you’re interested in helping to write the “transformer paper” and/or create the “ChatGPT moment” for Open Work.