Automated governance mechanisms in digital labour platforms: how Uber nudges and sludges its drivers

ABSTRACT

Using tools like machine learning algorithms, digital platforms raise new challenges to our understanding of control-governance dynamics in organisations. In this paper, we explore a unique governance mechanism; nudging – i.e. liberty-preserving approaches that steer people in particular directions – and provide exploratory findings that extend prior research in behavioural economics and organisational control-governance dynamics towards platform markets. We surveyed 166 Uber drivers to explicate the workings and effects of Uber’s good (i.e. transparent and easy to opt-out) and evil (i.e. obscure and misleading) nudges. Our findings suggest that while drivers are more satisfied with good nudges, these nudges do not make them more productive (i.e. increase their earnings-per-hour). Evil nudges, on the other hand, seem to have no effect on driver productivity. With experience, drivers learn to respond less to nudges (as they may realise that Uber’s nudges do not seem to increase their productivity). We extend the platform governance literature by highlighting whether and when nudges could influence drivers by creating false expectations. Our exploratory approach highlights new possible boundary conditions for the traditional theories, for example, Herzberg’s hygiene-motivation theory that, while differentiating hygiene factors from motivating factors, do not have the level of specificity to show the effects we discover here.

kEYWORDS:

• Platform governance
• nudging
• algorithmic management
• Uber

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/13662716.2022.2086450

Notes

¹ In the U.S., around 60 million precarious workers contribute $1.5 trillion in GDP. It is estimated that half of the U.S. labour force will be formed of gig workers by 2030 (Estes 2020).

² According to Thaler (2015), a nudge that fails to comply with one or more of these 3 principles qualify as an evil nudge: 1) All nudging should be transparent and never misleading. 2) It should be as easy as possible to opt out of the nudge, preferably with as little as one mouse click. 3) There should be good reason to believe that the behaviour being encouraged will improve the welfare of those being nudged.

³ Uber’s value creation depends entirely on having a driver quickly available around the customer.

⁴ Non-employees are distinct from workers in the sense that no specific governance arrangement could be embodied in written/formal contractual agreements as well as in unwritten/informal agreements (e.g. ‘self-bonding’ sunk costs, ‘social contracts’ implying a commitment to respecting and upholding social norms) (Argyres and Liebeskind 1999; Uzunca, Sharapov, Tee, 2022). As customers and machine learning algorithms replace managerial oversight and power, managing non-employees in platform markets requires different control and governance practices (Kellogg, Valentine, and Christin 2020; Stark and Pais 2020).

⁵ We would like to thank an anonymous reviewer for prompting us to better explain the hygiene-motivation theory.

⁶ A la Herzberg (1966), we could differentiate here between satisfiers and dissatisfiers where the first two criteria (transparent and easy to opt-out) qualify as motivating factors (factors which contribute to job satisfaction), whereas the last criterion (improve the earnings per hour of drivers) qualifies as a hygiene factor that contributes to job dissatisfaction.

⁷ Chen et al. (2019) estimated that Uber drivers earn more than twice the surplus they would in less-flexible arrangements. Similarly, Hall and Krueger (2018) examined survey evidence and documented that drivers cite flexibility as a reason for working for Uber and that many drivers report that Uber is a part-time activity, secondary to their regular employment.

⁸ Hence there are two main differences between fast forward and acceptance rate: 1) drivers who do not maintain their acceptance rate above 80% are threatened of being sacked out of the platform, whereas drivers who (temporarily) turn off the fast forward function will not experience any negative consequences. 2) Fast forward function allows drivers to complete more rides, which is in their interest, while keeping up a high acceptance rate is mostly in the interest of Uber and not necessarily in the interest of the drivers.

⁹ Good nudges might steer drivers towards valuing non-monetary incentives. For example, collecting badges or compliments, such as ‘cool car’, ‘entertaining driver’, or ‘awesome music’ (see Figure 2 for an overview of these badges) might not benefit drivers in terms of additional income. One can think how these are good nudges as by definition they do not serve drivers’ interest, which is a criterion for a good nudge. However, in terms of driver welfare, these nudges motivate drivers by increasing their satisfaction when driving for Uber. In this sense, drivers not only have monetary interest but a general welfare gain from these nudges. That is why badges and alike qualify for good nudges.

¹⁰ On the other hand, UberPop – Uber’s unlicensed driver service in Europe – has been banned in most European countries, and the practice of becoming an Uber driver varies widely within the continent (Uzunca and Ozcan 2022).

¹¹ We explicitly requested in posting our survey that only Uber drivers fill in our survey.

¹² Most drivers went to college, but did not obtain a degree (34%), or obtained a bachelor’s degree (29%). The remaining drivers completed high school (13%), had an associate degree (12%), or a master’s degree (10%). 3 drivers (2%) drivers did not complete high school.

¹³ Results are robust to logged number of rides drivers completed as an alternative dependent variable.

¹⁴ Experienced drivers are defined as having spent above the median number of months (21.4 months) since the start of driving for Uber.

¹⁵ According to a 2019 survey (https://therideshareguy.com/uber-driver-survey/) conducted by the industry expert Harry Campbell – better known as ‘The Rideshare Guy’ – 54% of Uber drivers are over the age of 50, and about a quarter are 61 or older. Another survey by SurveyMonkey Intelligence found that more than half of drivers are at least 40 years old. Therefore, the average age of 45.3 in our sample could be regarded as representative.

¹⁶ In the same survey by ‘The Rideshare Guy’, it is stated that ‘Only 18.8% of drivers are female,’ which also corresponds with our distribution of driver gender.