ArticlePDF Available


In the last decade, there has been a growth in, what we call, digitally mediated workplaces. A digitally mediated workplace is one where interactions between stakeholders are primarily managed by proprietary, algorithmically managed digital platform. The replacement of the relationships between the stakeholders by the platform is a key feature of these workplaces, and is a contributing factor to the decrease in contractual responsibilities each stakeholder has to one another. In this paper, we discuss some of the ways in which this structure and lack of accountability serves as a root of, or at least an enabler to, the realization of biases in the ridesharing application Uber, a digitally mediated workplace.
The Roots of Bias on Uber
Benjamin V. Hanrahan, Ning F. Ma, Chien Wen Yuan
College of Information Sciences and Technology at the Pennsylvania State Univer-
sity, USA,,
Abstract. In the last decade there has been a growth in digitally mediated workplaces.
That is, a workplace that is primarily mediated via an often propriety, algorithmically man-
aged, and where the majority of interactions between stakeholders take place. In fact the
replacement by the platform of the relationships between stakeholders is a key aspect of
these workplaces, and the root of the decrease in contractual responsibility each stake-
holder has to the others. In this paper we are particularly interested in how a particular lack
of accountability, biases, play out on Uber, a digitally mediated workplace, the ridesharing
1 Introduction
Recently, there has been a growth in digitally mediated workplaces, which is par-
tially defined by the particular stakeholder structure that they rely on: a platform
creator, who is responsible not only for defining and implementing the functionality
for the platform, but also the policies around the workplace that the platform imple-
ments or supplements; a worker, who uses the platform to find, claim, and obtain
remuneration from jobs; and a client, who uses the platform to procure and pay for
labor. This structure is instantiated by a number of different platforms for a num-
ber of different purposes, e.g.: Amazon Mechanical Turk (AMT), where the worker
is part of the crowd and the client requests work from the crowd; Fiverr, Upwork,
or Elance, which are primarily for freelancers to sell their services to clients; and
Uber (the focus of this paper), where the worker is the driver and the client is the
A key aspect of a digitally mediated workplace is that the, usually proprietary,
platform (e.g. AMT, Fiverr, Ola, TaskRabbit, or Uber) replaces much of the rela-
tionship between the worker and client, or the employee and employer. This also
drastically alters, if not eradicates, the contractual responsibilities of each stake-
holder to each other and reduces the level of accountability all around (sometimes
discussed as algorithmic accountability (Lustig et al., 2016; Lee et al., 2015; Wa-
genknecht et al., 2016)). In this paper, we are interested in a particular aspect of this
accountability, protection against bias for the worker and the client.
There is already evidence that bias and discrimination are having a demonstrable
impact on the stakeholders of these platforms (Hann´
ak et al., 2017; Edelman and
Luca, 2014). However, existing work has looked more at the existence of bias and
less about how biased decisions are performed on or via these platforms. To begin
to investigate how to design to avoid bias more broadly on these platforms, we first
need to look at how bias is specifically occurring and what its roots might be on
a specific platform. The specific platform that we report on in this paper is Uber,
a Ridesharing or Transportation Network Company, where passengers obtain rides
from the drivers. Drivers must use their own cars and obtain rides via the Uber app.
We argue that, while Uber certainly is not wholly a digital workplace, it is a
digitally mediated workplace. That is, while there are certainly face-to-face inter-
actions between the driver and the passenger, these exchanges are arranged via the
Uber app and the consequences of the interactions are mediated by the app. So
Uber serves as an interesting mixed-setting for a digitally mediated workplace, as
consequences of face-to-face interactions are both captured and propagated through
the digital platform.
This means, that while there are face-to-face interactions between the worker
and client, whatever rating they give each other is mediated solely through digital
means. That is, there is no human in the loop to take different factors into account
or impart a level of flexibility or subjectivity to the process. As these ratings have
a real impact on both the driver and passenger’s ability to provide and procure ser-
vices, this opens up an avenue for unfettered biased judgements that are propagated
by the platform (Mcgregor et al.). To best illustrate our point, we provide this specu-
lative comparison. In an existing, more traditional taxi service, if a passenger would
like to make a biased complaint they must call a supervisor, or at the very least a
representative of the taxi service. During this call, there is a likelihood that the su-
pervisor will uncover or detect the bias due to the existing relationship between the
supervisor and the driver, in addition to the supervisor’s judgement as to the validity
and veracity of the complaint. So there is at least some level of human mediation
when fielding complaints. Contrast this to a biased complaint on Uber, where the
only signal of the complain might be a rating, where all of the nuance and reasoning
behind that are not interrogated or even captured by the system. This biased judge-
ment then propagates throughout the system since that rating is used by the system
and users to determine which driver to send to a job.
In this paper, we draw from a similar methodology as Martin et al. (2014) where
we examine what discussions Uber drivers are having regarding bias online. As
we argue that Uber is a digitally mediated workplace, online forums are a place
where the shop talk happens, similarly to Turkers. For this paper, we looked to see
whether or not and how the drivers discussed the effect of biases on them or even
their own biases. We report some of our preliminary findings on how biases bear
out both by and towards drivers on Uber and what the role of the platform is. In this
way, we are beginning to look at how the same phenomena that led to protections
for workers and customers in traditional workplaces are reoccurring in digitally
mediated ones. Analyzing the practice of bias is the first step towards designing
similar functionality that govern these digitally mediated workplaces to the policies
that govern tradition workplaces.
2 Related Work
In this section we review research into digital mediation of work and how biases
may be being enacted in the workplace.
2.1 Peer-to-peer platforms and technology mediation
Beyond technological tools that mediate work like email (Hinds and Kiesler, 1995),
instant messaging (Isaacs et al., 2002), or social network site (Dimicco et al.), peer-
to-peer (P2P) platforms like Uber, Lyft, or Ola are digitally mediated workplaces
where workers manage their tasks and negotiate transactions with their clients both
online and offline. While the task is completed offline, such as the driver sending
the passenger to a destination and potentially engaging in social interactions with
each other along the way, many practices are structured by technological features
and computational algorithms of the platforms. Automated dispatch systems use
genetic or optimization algorithms and devices with built-in GPS to match drivers
with passengers in real time based on geo-locations (Karande and Bogiri, 2015;
Rawley and Simcoe, 2013). Fares and payment rates are set based on locations,
times of the day (e.g., higher in rush hours), and the services requested (e.g., single
ride or shared ride). In addition to real-time data, Uber assigns work to drivers and
allows passengers to request services based on the historical data, namely the rating
system on the platform (Ahmed et al., 2016).
Much previous work has investigated issues revolved around such computing
systems and algorithms, and their influences on users. Automated dispatch sys-
tems may deploy drivers to move outside their familiar geographic areas (Hsiao
et al., 2008). While this allows drivers to acquire information about some poten-
tial hotspots, it also demands drivers to develop temporal and spatial knowledge.
Devices with GPS systems shape drivers’ wayfinding and navigation skills and po-
tentially change the social dynamics of the riding processes between drivers and
passengers (Girardin and Blat, 2010; Hsiao et al., 2008). With their influences on
practices and work revolved around the P2P platforms, the most prominent issue
with these algorithms and systems is the lack of transparency to users (Lustig et al.,
2016). Despite the invisibility and inaccessibility, users still have to make sense of
how to interact with the systems in order to manage their work (Lee et al., 2015),
rely on the digital infrastructure to quantify their work and develop their account-
ability using the rating system (Scott and Orlikowski, 2012), or deal with potential
offline consequences like the uncertainty of finding next customer by taking request
from the dispatch system Ahmed et al. (2016).
Designed to collect data to facilitate coordination or even prediction of human
work, computing systems and algorithms are often valued for their instrumental
functions. Given these identified issues, computing systems and algorithms may
not be posed as neutral and objective as they may seem (Kneese et al., 2014) . It is
possible that the digital infrastructure imposes and renders biases, intentionally or
unintentionally, against users (Wagenknecht et al., 2016) . In this study, we com-
plement prior work by exploring and identifying some existing biases. We include
experiences from both workers and clients, using Uber as our target platform, in an
attempt to draw a holistic picture revolved around this issue.
2.2 Biases at workplace
Biases usually refer to stereotypical generalizations based on sociodemographic
or physical characteristics about certain groups that are assigned to the individual
group members. Previous research reported gender biases (Heilman, 2012), ageism
(Rupp et al., 2006), racial biases (Rosette et al., 2008), or weight bias (Rudolph
et al., 2009) at workplace. These biases are associated with inequality in employ-
ment decisions, career advancement, performance expectation, workload, overall
evaluations, etc.
While these biases are prevalent in physical workplace because the characteris-
tics and attributions are visible and obvious to elicit implicit or explicit biases, they
do not disappear even if the work is mediated. Research reported that biases also
took place on technological platforms. For example, workers on TaskRabbit used
geolocations to evaluate whether to accept a task and were found that they tended
to avoid distant and less well-to-do areas (Thebault-Spieker et al., 2015). On the
other hand, clients may also choose workers from these P2P platforms based on
their gender and race no matter the tasks are completed in physical or virtual con-
texts (Hann´
ak et al., 2017). Workers have to have adequate equipment like bank
accounts, smartphone with built-in GPS or in UberBlacks case, a fancy car, to be
able to provide services (Kasera et al., 2016).
In addition to biases rendered by sociodemographical and physical factors, we
argue that on the digitally mediated workplace, these biases could potentially be
reinforced and propagated by the digital infrastructure.
The rating system on Uber represents a record of drivers’ work performance
and is used to evaluate their eligibility to receive service requests. However, there is
no clear metric, such as driving skills, safety concerns, or decision-making strate-
gies about picking up routes, as to how the performance is evaluated. Instead,
drivers may have to engage in “emotional labor, in which they need to quickly
build “micro-relationships” that make passengers feel good so as to get good rat-
ings (Nardi, 2015; Rogers, 2015; Mcgregor et al.). Such emotional labor is easily
influenced by random factors and the efficacy and accuracy of the rating system
may benefit from a more holistic evaluation (Lee et al., 2015).
In addition, while racial and gender biases are suggested to be mitigated through
Uber’s matching algorithm, Mcgregor et al. pointed out that the algorithm actually
denies users ability to choose their desirable drivers or passengers and therefore
deepens the negative effect of expected homophily for both drivers and passengers.
The consequence may be a lowered rating. On Uber platform, drivers usually have
to respond to requests within 15 seconds without knowing the destination and ex-
pected fare. In order to avoid deactivation from the platform, Uber drivers often do
not have sufficient time for decision-making (Rosenblat and Stark, 2016) and have
to deal with offline consequences reinforced by the platforms (Ahmed et al., 2016).
In our study, we explore several different occurrences of biased practices and
judgements that are either enabled by the digital infrastructure or rooted in an aspect
of it.
3 Method
In investigating if and how Uber drivers discuss bias in the workplace, we borrowed
heavily from the approach taken by Martin et al. (2014) in their study of Turk-
ers’ issues and concerns. We focused on the most popular forum for Uber drivers, The primary way that we differ from Martin et al. (2014),
is that we were interested in a specific topic and did not let all of the topics that
concern Uber drivers emerge from out study. That said, we still took an exploratory
approach to our investigation around bias in the workplace, looking at all forms and
instances, e.g. not just biases on the part of passengers, but also biases expressed
by the drivers on the forums.
UberPeople is a forum that is run by drivers for drivers. The current users are
from major cities around the world with most of active members located within
the US. The forum is divided into many sections, the ones that we looked most
closely at were community related: Advice,Stories,People, and Complaints. The
Advice section is the most active section, closely followed by the Complaint section,
the other sections Stories and People, have significantly less activity. The primary
source of the content in this paper are from the Complaint section.
For two months, we have started to collect content from the various posts on the
forum and gathered threads posted between January 2015 - February 2017, which
are relevant to the bias theme. From the collected threads, in the preliminary pa-
per we report on 16 selected threads that represent a range of biased practices and
scenarios in the workplace. To gauge how broad and valid the content of the dif-
ferent posts were, we looked at the responses by the community. For instance, if
a user wrote a post making an uncommon, potentially outrageous claim, then the
community would respond in kind. That said, outrage at a claim of bias is not un-
common so we to avoid false negatives. However, if the community is supportive
and is in agreement this is a strong sign that a phenomena is valid. For any threads
that contains a mix of opinions on the part of the forum users, we present both sides
of the argument. All the selected posts and threads are categorized as being rooted
in either a lack of transparency or lack of recourse. While presenting the different
themes that emerged we make note of whether or not these are biases impacting
drivers or passengers.
4 Findings
In our reading of the forum we saw a number of themes emerge around the discus-
sion of biases on Uber, some were enabled by the platform, some where exacted by
the platform itself, and were clearly due to the behaviors by the different stakehold-
ers. That is, there are some biases that are seen as inherent in the design of the Uber
marketplace and tool. While there are other biases that may be propagated or sup-
ported by the system unwittingly, but clearly originate from one of the stakeholders
and are clearly directed at another specific stakeholder. Somewhat surprisingly to
us, we found a diverse set of biases, that is, while we expected – and did – see bi-
ases that impacted the drivers (drivers were after all the primary users of the forum),
we also saw discussions about biases aimed towards passengers by both the drivers
and the platform structure. We saw two main roots to the perception or practice of
biases: the lack of transparency, this manifested mostly in the rating system; and
the lack of recourse, there was no clear way to take recourse against what drivers
saw as biased judgements, so they developed strategies which themselves contained
4.1 Biases Rooted in a Lack of Transparency
One of the frustrations that drivers had with Uber’s rating system is that it is not
terribly transparent with respect to passengers’ ratings, which was especially con-
cerning when the drivers had received low ratings.
The reason why we need to know who rated to be able to fix any issue ... This system
will make riders more accountable before they ruin someones life. - F1
At times, this lack of transparency leads drivers down a path of suspicion. It is
hard for the drivers to know what exactly they did to deserve a poor rating and they
begin to speculate about a variety of reasons. When drivers belong to a minority
and are receiving low ratings for reasons that are unknown to them, they begin to
speculate – with ample reason at times – that it is related to a particular bias.
4.1.1 Racial
Drivers are clearly aware of the potential for biased ratings, as well as the inability to
know whether or not bias has influenced their ratings. Drivers are certainly worried
that biased ratings might be impacting them.
If I were black and got deactivated I’d be screaming from the hilltops about racism.
It’s probably THE best argument against the rating system there is... Ageism is abso-
lutely a factor too. But if you are an older black male I would say it’s worse... But the
bottom line is the ratings are unfairly applied. It probably depends on the area and the
demographics of the customer base as to HOW they are unfairly applied. But anyone
who thinks race isn’t a factor (and ageism and sexism) in any system is deluded. - F2
One user believed that they were suffering from biased ratings, which was par-
ticularly problematic as they had just started and were in danger of being deacti-
This is my 4th day driving. My rating now stands at 4.64... I just can’t figure out why
my rating are borderline deactivation level. This is crazy. I’m curious, especially to
hear from other young(ish) black male drivers if they are constantly on the borderline
as well. I hate even having to bring up this topic, but honestly I don’t know what else
I could even be doing to bring my rating up. - F3
Conversations around biases, particularly racism, seem to become contentious
fairly quickly on the forum (similar to other venues). When the issue is specifically
called out by a user, passionate voices fall on both sides of the issue. Some minimize
and deride the claim of bias:
Every bad thing in your life that happens to you is racially motivated. “The man” is
out to get you. - F4
Others provide support and counter other members to defend the original poster:
You can talk all the sh!t you like, I am a black man in America, I see, hear and
experience racism on a weekly basis. - F5
4.1.2 Other Biases
There were other biases related to language that one driver claimed to have noticed.
I’ve noticed a number of posts by poor-English speakers about bad ratings. That’s
probably one of the most difficult biases to overcome. - F6
One user hypothesized that all manner of biases are probably at play in the rating
Of course the crowd-sourced rating system is racist. Probably sexist and ageist too.
Ugly people get lower ratings than attractive people too. - F7
It seems clear that the lack of transparency behind the reasoning for passengers’
ratings is opening the door to biased ratings that are unfettered by the system. At
the very least, this lack of accountability is leading to a lot of suspicion. Drivers
even speculated that Uber assigns certain types of people to certain types of areas:
I think as much as possible Uber tries to send us black drivers into the “hood”.... To
pick up black passengers.... This morning I was at the air port the 3rd one to go
out....when I get a ping...I look at my phone, and see the pax is 25 min away and has
a very ethnic specific name - F10
Although this was met with skepticism from other drivers, and other drivers
encouraged the driver to be more selective about what types of neighborhoods or
distances that they traveled for their passengers.
4.2 Strategies in Response to Perceived Bias
While there is evidence on the forums that drivers are impacted by the biases of
passengers, there is also evidence of the various strategies that drivers had developed
in response.
Passengers themselves are not immune to the biases of the drivers either. The
biases that we saw on the part of the drivers, were surprisingly rooted in practices
that drivers had enacted as a strategic response to the perception of passenger biases.
4.2.1 Ignore
One of the more innocuous strategies that drivers suggested in an earlier example
to F3, was to just tolerate the bias as a part of doing business. They advised not
to worry about it as cases of bias are absorbed by the majority of good, decent
passengers and as time when on these incidents had less and less impact on their
overall rating.
It only takes one rider to dent your rating when you’re new. I wouldnt worry too much
just yet. - prk
Seriously, do not worry about your rating this early in the game. I get the exact same
BS feedback you got at 4.92 ratings after 500 plus rides. - F11
Simply to tolerate this intolerance is anathema to the zero tolerance policy to
which Uber subscribes 1.
4.2.2 Retaliation, Protest
In one case of a driver being frustrated by receiving poor ratings for inscrutable
reasons, a driver decided to take a protest action of giving any and all passengers
that they gave a ride to on that day a poor rating.
Ok. So since Uber doesnt let us know who give us a bad rating and leave us guessing.
I decided to punish all riders of the day if my rating goes down .01 point. ... I think
we have the right to know who rate us bad and the reason. Otherwise i will use this
method. I know it wont matter. But when the rider check their ratings they will see
how it dipped down too. - F1
In another thread discussing the effect of biased ratings on the drivers, the con-
versation turned towards speculation about ‘certain areas’ and ‘stupid biases’ being
the source of poor ratings. One user had taken a similarly oppositional practice of
awarding high ratings only to exceptional passengers.
Im done worrying about riders so much. If you work around certain areas. Youll
realize your rating drops even if you keep the cleanest car and is the best driver. Now
the pax needs to amuse me to get over 4 stars. Stupid Biases and complexes really get
in the way. - F14
4.2.3 Avoidance
The instances of driver bias towards passengers mostly happened in how the drivers
tried to avoid certain areas or types of passengers.
One example, is a driver who after a bad experience with passengers from the
Black Entertainment Television awards, experienced a dip in their rating and came
to this conclusion:
I’m not ignorant of the racial tensions in this country right now. I’m sure there’s some
real animosity. I think there’s something about Rap too that brings out the hate. Now
when I see a group of black guys I’m automatically going to just hit cancel. I hate
saying that too because I love my black friends but what are you going to do. - F9
In this same thread, other drivers provided numerous counter examples where
they had positive experiences with African American passengers. Clearly, there is
the potential for drivers’ biases to impact passengers’ ability to procure a ride.
A different driver had another set of much more blatantly racist complaints about
a different group of riders, framing them as others that even inhabit a different world
of sorts.
1 They do not know this is a ride-sharing. They treat you like a low-educated, no-skill
cab driver. 2 They intentionally make you wait for up to 5 minutes 3 They ask you
drive up to the front door even they live in an apartment complex...4 Most of them
have very strong body odors... 5 Most of their rides are a $4 trip including pick up
from or go to the Indian grocery store or Indian restaurant...7 They never tip...8 They
gave you wrong directions and blame you taking the longest route from point A to
B. 9 They give you lower rating too. In their world, a 5star is impossible and never
exists. - F8
These avoidance strategies where “experienced” drivers make use of Uber’s can-
cellation policies which provides a loophole to drivers who want to avoid passengers
and suffer few if any consequences. These strategies do have a negative impact on
the passengers, which can be seen in one of the rare instances of a passenger posting
to the forum.
This guy wasted my time (which apparently was very precious in that span), didn’t
answer my calls, THEN had the nerve to charge me a cancellation fee! Isn’t there
some way to rate this guy as unprofessional? I have his ID number. - F12
This passenger was canceled by the driver in a severe weather day. Due to the
app system design, the passenger was charged a fee while his/her trip was canceled.
This shows that there is at least a reciprocal avenue through which passengers can
also be impacted by drivers’ biases.
5 Discussion
When we set out to start this study, we expected to see drivers discussing the im-
pact of biases on themselves. What we were surprised by, was the candor with
they discussed their own biases and how these biases impacted passengers. One fo-
rum member felt that the various avoidance strategies that drivers used were being
reinforced by the various pricing strategies that Uber employs.
Uber has brought back redlining with its boost incentives. It is subsidizing the rides
of the well off, mostly white riders on the west side and leaving minorities and lower
income residents in Central LA and South LA with fewer drivers. Uber, ..., are the
ones responsible for ride share redlining ... - F13
Redlining is a practice that originates in more traditional taxi companies, where
the companies refused fares from low-income communities. This practice of taxi
companies was dealt with via legislation, but now seems to be reemerging on Uber.
5.1 Transparency
When biases are more apparent and obvious, Uber is able to take action. Such was
the case when a Raleigh, NC same-sex couple was kicked out of an Uber driver’s
car, their story was covered in the media and discussed later in the forum with mixed
voices. Uber released a statement and blocked the driver from giving rides on Uber.
However, the small instances of bias that we have seen evidence of, be it by either
drivers or passengers, are much more difficult to trace and take action on.
A great deal of the problem of detecting bias and preventing bias is rooted in the
lack of transparency in the rating system. This lack of transparency led to a lack of
accountability in terms of the giving of ratings, as well as the suspicion that biases
were having an effect on the drivers’ status. The biases that we saw in our study
were either directly enabled by the lack of transparency, or in direct response to the
presumption of bias due to this lack of transparency.
5.2 Design Implications
There are two preliminary design implications from our findings. First, we would
argue for a higher degree of transparency in the reasons behind low ratings. This
could take the form of pinging the author of the low rating for additional more qual-
itative feedback that the driver could take action on. Additionally, Uber can better
leverage the various data points that they are gathering about the ratings and inter-
actions between a particular passenger and different drivers, or between a particular
driver and different passengers. For instance, Uber may be able to track a certain
passenger’s reactions to different demographics and use this information to reduce
the weight of that persons ratings. Perhaps, the passenger could be confronted with
this perceived bias, as it may be implicit and not realized, so that they can act to
remedy their own bias.
6 Limitations
Our preliminary study has obviously limitations in the length of time that we have
been collecting data and the breadth of data that we have collected. That said, we
feel that we have several concrete examples of a relatively rarely discussed phe-
nomenon that map to the bias that other researchers have reported on these plat-
forms. We have also began to identify some of the strategies that drivers have taken
in response to perceived bias.
Ahmed, S. I., N. J. Bidwell, H. Zade, S. H. Muralidhar, A. Dhareshwar, B. Karachi-
wala, C. N. Tandong, and J. O’Neill (2016): ‘Peer-to-peer in the Workplace’.
In: Proceedings of the 2016 CHI Conference on Human Factors in Computing
Systems - CHI ’16. New York, New York, USA, pp. 5063–5075, ACM Press.
Dimicco, J. M., D. R. Millen, W. Geyer, and C. Dugan. ‘Research on the Use of
Social Software in the Workplace’.
Edelman, B. and M. Luca (2014): ‘Digital Discrimination: The Case of’. Harvard Business School, pp. 21.
Girardin, F. and J. Blat (2010): ‘The co-evolution of taxi drivers and their in-car
navigation systems’.
ak, A., C. Wagner, D. Garcia, A. Mislove, M. Strohmaier, and C. Wilson
(2017): ‘Bias in Online Freelance Marketplaces: Evidence from TaskRabbit
and Fiverr’.
Heilman, M. E. (2012): ‘Gender stereotypes and workplace bias’. Research in
Organizational Behavior, vol. 32, pp. 113–135.
Hinds, P. and S. Kiesler (1995): ‘Communication across Boundaries: Work, Struc-
ture, and Use of Communication Technologies in a Large Organization’. Orga-
nization Science, vol. 6, no. 4, pp. 373–393.
Hsiao, R.-L., S.-H. Wu, and S.-T. Hou (2008): ‘Sensitive cabbies: Ongoing sense-
making within technology structuring’. Information and Organization, vol. 18,
no. 4, pp. 251–279.
Isaacs, E., A. Walendowski, S. Whittaker, D. J. Schiano, and C. Kamm (2002):
‘The character, functions, and styles of instant messaging in the workplace’. In:
Proceedings of the 2002 ACM conference on Computer supported cooperative
work - CSCW ’02. New York, New York, USA, p. 11, ACM Press.
Karande, N. B. and N. Bogiri (2015): ‘Solution To Carpool Problems using Genetic
Algorithms’. International Journal of Engineering and Techniques, vol. 1, no. 3.
Kasera, J., J. O’Neill, and N. J. Bidwell (2016): ‘Sociality, Tempo & Flow: Learn-
ing from Namibian Ridesharing’. In: Proceedings of the First African Confer-
ence on Human Computer Interaction - AfriCHI’16. New York, New York, USA,
pp. 36–47, ACM Press.
Kneese, T., A. Rosenblat, and <.<!>boyd (2014): ‘Understanding Fair Labor
Practices in a Networked Age’. SSRN Electronic Journal.
Lee, M. K., D. Kusbit, E. Metsky, and L. Dabbish (2015): ‘Working with Machines:
The Impact of Algorithmic and Data-Driven Management on Human Workers’.
In: Proceedings of the 33rd Annual ACM Conference on Human Factors in Com-
puting Systems. pp. 1603–1612.
Lustig, C., K. Pine, B. Nardi, L. Irani, M. K. Lee, D. Nafus, and C. Sandvig (2016):
‘Algorithmic Authority: the Ethics, Politics, and Economics of Algorithms that
Interpret, Decide, and Manage’. In: Proceedings of the 2016 CHI Conference
Extended Abstracts on Human Factors in Computing Systems - CHI EA ’16. New
York, New York, USA, pp. 1057–1062, ACM Press.
Martin, D., B. V. Hanrahan, J. O’Neill, and N. Gupta (2014): ‘Being a turker’. In:
CSCW 2014. pp. 224–235.
Mcgregor, M., B. Brown, M. Gl¨
oss, and A. Lampinen. ‘On-Demand Taxi Driving:
Labour Conditions, Surveillance, and Exclusion’.
Nardi, B. (2015): ‘Inequality and limits’. First Monday, vol. 20, no. 8.
Rawley, E. and T. S. Simcoe (2013): ‘Information Technology, Productivity, and
Asset Ownership: Evidence from Taxicab Fleets’. Organization Science, vol. 24,
no. 3, pp. 831–845.
Rogers, B. (2015): ‘The Social Costs of Uber’. The University of Chicago Law
Review Dialogue, vol. 82, no. 85, pp. 85–102.
Rosenblat, A. and L. Stark (2016): ‘Algorithmic Labor and Information Asymme-
tries: A Case Study of Uber’s Drivers’. International Journal of Communication,
vol. 10, pp. 3758–3784.
Rosette, A. S., G. J. Leonardelli, and K. W. Phillips (2008): ‘The White standard:
Racial bias in leader categorization.’. Journal of Applied Psychology, vol. 93,
no. 4, pp. 758–777.
Rudolph, C. W., C. L. Wells, M. D. Weller, and B. B. Baltes (2009): ‘A meta-
analysis of empirical studies of weight-based bias in the workplace’. Journal of
Vocational Behavior, vol. 74, no. 1, pp. 1–10.
Rupp, D. E., S. J. Vodanovich, and M. Crede (2006): ‘Age Bias in the Workplace:
The Impact of Ageism and Causal Attributions1’. Journal of Applied Social
Psychology, vol. 36, no. 6, pp. 1337–1364.
Scott, S. V. and W. J. Orlikowski (2012): ‘Reconfiguring relations of accountability:
Materialization of social media in the travel sector’. Accounting, Organizations
and Society, vol. 37, no. 1, pp. 26–40.
Thebault-Spieker, J., L. G. Terveen, and B. Hecht (2015): ‘Avoiding the South Side
and the Suburbs’. In: Proceedings of the 18th ACM Conference on Computer
Supported Cooperative Work & Social Computing - CSCW ’15. New York, New
York, USA, pp. 265–275, ACM Press.
Wagenknecht, S., M. K. Lee, C. Lustig, J. O’Neill, and H. Zade (2016): ‘Algorithms
at Work: Empirical Diversity, Analytic Vocabularies, Design Implications’.
... At first, the community was relatively optimistic about the promises of the peer-to-peer economy, which was initially framed as the "sharing" economy, a term that highlights the positive aspects of these exchanges [6][7][8]. However, as time passed that optimism faded and the community became more wary of these peer-to-peer economies, which is now more often framed as the "gig" economy, a term that highlights the instrumental, exploitative, and negative aspects of these exchanges [23,41,43]. That said, there are still instances of the peer-to-peer economy that seem to embody some of the initial optimism of the sharing economy, e.g., timebanking [46,47], and there have been efforts to build systems to support these sharing practices [48]. ...
... Part of this dynamic is manifested in the rating system [43], where drivers have experienced and exercised bias against riders from ride-hailing platforms [23,49]. ...
... Reflecting on the benefits and promises that ride-sharing activities could bring about [27,31], as well as the current challenges in the ride-hailing services [23,26], we intend to unpack the benefits and challenges of ride-sharing in a unique carpooling context and expand the current understanding of sharing economy with a more granular perspective. ...
Full-text available
In recent years, the peer-to-peer economy has grown exponentially, particularly in the ride-sharing sector. This growth has been accompanied by a muddying between the sharing and gig economy, and it has become unclear when an activity is sharing a resource vs. providing a service. To unpack this difference, we studied two successful carpooling groups (university students traveling home and commuting among professionals), which we contrast with previous literature on ride-hailing apps (e.g., Uber). The two communities that we studied differ in that: professionals, had more routine ride-sharing needs based on their commute; and students, arranged rides to return home for school breaks or long weekends. We detail how common needs and backgrounds impacted how carpoolers treated each other. Leveraging these findings, we outline design paths for both the sharing and gig economies to better realize the ideas of the sharing economy.
... This paper contributes to ongoing conversations in CSCW about transaction costs and power asymmetries in the on-demand economy by providing scalable recommendations for companies that use the management are sometimes portrayed as a way to avoid human bias against workers, in the case of platforms like Upwork there may be significant bias because clients have the option to select the specific freelancers themselves and rate them, which has been shown on other platforms to be biased against Black people and women [13]. On platforms in which both clients and freelancers can be rated, this bias may go both ways [14]. Furthermore, D'Cruz and Noronha found that Indian workers on Upwork had a different experience than those typically reported in the literature on Western workers: "Aversive racism from clients and fellow freelancers, calling into question workers' competence, bid amounts and remuneration rates, accompanied by divergences linked to ethos, language and time, bring a negative tenor to work-related Work organisation, interactions and make it more difficult to complete tasks that are already complicated by their virtual and often asynchronous forms" [3]. ...
... Researchers have called for more study of how platforms control the information that both sides of the market have access to [1]. Some researchers have suggested design interventions for platforms, primarily for Amazon Mechanical Turk (e.g., [14,18,33]). In summary, change can happen through laws, corporate self-regulation, the organizing efforts of freelancers, and redesign of on-demand platforms; and we believe all these avenues for change should be pursued. ...
... Therefore, we suggest that employees' reviews could be tied to the rating that the freelancer gives them. Although, as noted in Hanrahan et al. [14], ratings can be biased against women and racial minorities, so the implementation of our suggestion must be carefully considered. Fourth, provide employees with information about when it is appropriate to use the freelance economy versus other types of labor. ...
Full-text available
Corporations are increasingly empowering employees to hire on-demand workers via freelance platforms. We interviewed full-time employees of a global technology company who hired freelancers as part of their job responsibilities. While there has been prior work describing freelancers' perspectives there has been little research on those that hire them, the "clients", especially in the corporate context. We found that while freelance platforms reduce many administrative burdens, there are number of conditions in which using freelance platforms in a corporate context creates high transaction costs and power asymmetries that make it difficult for clients to negotiate work rights and responsibilities. This leads corporate employee clients to feel "stuck in the middle" between their employer, the platform, and the freelancer. Ultimately, these transactions costs are a potential barrier to wider adoption. If corporations want to leverage the value of the freelance economy then better guardrails, guidelines, and perhaps even creative technology solutions will be needed.
... A study by Hanrahan et al. (2017) relates bias to the rating system of Uber. After a ride, the driver and passenger rate each other. ...
... To be sure, scholars and activists have suggested ways in which some of the above concerns could be addressed. Hanrahan et al. (2017) argue that platforms like Uber can become what they call a 'vehicle of basis' due to the lack of transparency and accountability of the rating system. They suggest two design strategies to overcome this: (1) a higher degree of transparency in rating and (2) tracing whether certain users systematically give biased ratings and lower the weight of their ratings in the overall rating score. ...
... The mentioned study by Hanrahan et al. (2017) is particularly relevant in this respect. It shows how the rating system has become a vehicle of bias. ...
Full-text available
In this paper, we argue that the characteristics of digital platforms challenge the fundamental assumptions of value sensitive design (VSD). Traditionally, VSD methods assume that we can identify relevant values during the design phase of new technologies. The underlying assumption is that there is only epistemic uncertainty about which values will be impacted by a technology. VSD methods suggest that one can predict which values will be affected by new technologies by increasing knowledge about how values are interpreted or understood in context. In contrast, digital platforms exhibit a novel form of uncertainty, namely, ontological uncertainty: even with full information and overview, it cannot be foreseen what users or developers will do with digital platforms. Hence, predictions about which values are affected might not hold. In this paper, we suggest expanding VSD methods to account for value dynamism resulting from ontological uncertainty. Our expansions involve (1) extending VSD to the entire lifecycle of a platform, (2) broadening VSD through the addition of reflexivity, i.e. second-order learning about what values to aim at, and (3) adding specific tools of moral sandboxing and moral prototyping to enhance such reflexivity. While we illustrate our approach with a short case study about ride-sharing platforms such as Uber, our approach is relevant for other technologies exhibiting ontological uncertainty as well, such as machine learning, robotics and artificial intelligence.
... On ride-sharing platforms, drivers go to great lengths to ensure that they are competitive with the experiences that other drivers are providing -activities which are not accounted for or remunerated [48]. Another knock-on effect of this asymmetry is illustrated by how bias is practiced on the various platforms [25,51,56,24]. For example, race and gender were found to be correlated with performance met-rics (which impacts opportunities for new assignments) [24,51], or even -in the case of AirBnB -is associated with lower remuneration [14]. ...
... For example, race and gender were found to be correlated with performance met-rics (which impacts opportunities for new assignments) [24,51], or even -in the case of AirBnB -is associated with lower remuneration [14]. In part, these biases are rooted in how the functionality and anonymous structure of these platforms lead to a lack of accountability in the relationships that they purport to support [25]. ...
... "Redlining" is a practice that originates in more traditional taxi companies where the companies refused fares from lowincome communities. This practice was dealt with legislation back then, but now seems to be reemerging on Uber [25]. ...
Conference Paper
Full-text available
Uber is a ride-sharing platform that is part of the 'gig-economy,' where the platform supports and coordinates a labor market in which there are a large number of ephemeral, piecemeal jobs. Despite numerous efforts to understand the impacts of these platforms and their algorithms on Uber drivers, how to better serve and support drivers with these platforms remains an open challenge. In this paper, we frame Uber through the lens of Stakeholder Theory to highlight drivers' position in the workplace, which helps inform the design of a more ethical and effective platform. To this end, we analyzed Uber drivers' forum discussions about their lived experiences of working with the Uber platform. We identify and discuss the impact of the stakes that drivers have in relation to both the Uber corporation and their passengers, and look at how these stakes impact both the platform and drivers' practices.
... Gig-workers appreciate autonomy and flexibility in their activity, but this is counter-balanced by several issues. These are: platformic or algorithmic management [7,8,18,21]; incentivization to work for some time and place [21,22]; the lack of transparency on the consequences of rating [5]; and unpredictable task assignment [5,12]. The algorithmic management implies that the "app" defines all aspects of the work: task assignment and waiting times before new tasks, reviews, work assessment and pay. ...
... Gig-workers appreciate autonomy and flexibility in their activity, but this is counter-balanced by several issues. These are: platformic or algorithmic management [7,8,18,21]; incentivization to work for some time and place [21,22]; the lack of transparency on the consequences of rating [5]; and unpredictable task assignment [5,12]. The algorithmic management implies that the "app" defines all aspects of the work: task assignment and waiting times before new tasks, reviews, work assessment and pay. ...
... Part-Time Ride-Sharing: Recognizing the Context in which Drivers Ride-Share and its Impact on Platform Use 1 INTRODUCTION Ride-sharing companies have, in some significant ways, reshaped the structure and practice of ride-hailing work. This reshaping is part of a more impactful trend in the gig-economy, where both journalists [1,8] and academics have discussed the various challenges [17,22,27,44] of ride-sharing. As a result of this re-shaping, some have highlighted the danger of the 'playbor' trend [43] -particularly the blurring distinction between work, hobby, and volunteerism [27] in the digitally mediated workplace. ...
... Due to the opaqueness of the platform designs, and the difficulty in properly operationalizing the many nuances and details of human interaction, Rosenblat et al. argued that these platforms create or facilitate asymmetries in information and power, where these asymmetries regularly favor the corporations' interest [42]. One example of this asymmetry is the rating system in ride-sharing platforms, where workers are known to both experience and practice bias [17,41]. ...
Full-text available
Ride-sharing companies have been reshaping the structure and practice of ride-hailing work. At the same time, studies have been showing mixed driver experiences on the platform while many of the drivers are working part-time. In this research, we seek to understand why drivers on this platform are working part-time, how this impacts their view of the platform, and what this means for more accurately evaluating the design of these platforms. To investigate this question, we focused on situating ride-sharing in the lives and constellation of gigs that drivers maintain. We collected 53 survey responses and conducted 10 semi-structured interviews with drivers to probe these questions. We found that the extent that drivers categorize themselves as part-time is less about the number of hours worked and more about how dependent they are on ride-sharing income. The level of this dependency seemed to heavily influence how they interacted with the platform and their attitudes towards difficulties faced. It seemed to us that in some ways that the design or functioning of the platform almost pushed users towards working part-time. We discuss the importance of taking these different types of workers and their situations into consideration when evaluating the design and usability of these platforms.
... Likewise, until survey respondents actually experience self-driving vehicles, they might pay relatively little attention to availability, privacy, and fairness, relative to the more salient issues of safety (Kaur & Rampersad, 2018). For example, the algorithms for sending a vehicle to pick up a passenger might be unfair and biased toward certain socioeconomic groups, similar to the biases that have emerged in Uber's ride request algorithms (Hanrahan, Ma, & Yuan, 2017), and algorithms more generally (Courtland, 2018). ...
Objective This study examined attitudes toward self-driving vehicles and the factors motivating those attitudes. Background Self-driving vehicles represent potentially transformative technology, but achieving this potential depends on consumers’ attitudes. Ratings from surveys estimate these attitudes, and open-ended comments provide an opportunity to understand their basis. Method A nationally representative sample of 7,947 drivers in 2016 and 8,517 drivers in 2017 completed the J.D. Power U.S. Tech Choice Study SM , which included a rating for level of trust with self-driving vehicles and associated open-ended comments. These open-ended comments are qualitative data that can be analyzed quantitatively using structural topic modeling. Structural topic modeling identifies common themes, extracts prototypical comments for each theme, and assesses how the survey year and rating affect the prevalence of these themes. Results Structural topic modeling identified 13 topics, such as “Tested for a long time,” which was strongly associated with positive ratings, and “Hacking & glitches,” which was strongly associated with negative ratings. The topics of “Self-driving accidents” and “Trust when mature” were more prominent in 2017 compared with 2016. Conclusion Structural topic modeling reveals reasons underlying consumer attitudes toward vehicle automation. These reasons align with elements typically associated with trust in automation, as well as elements that mediate perceived risk, such as the desire for control as well as societal, relational, and experiential bases of trust. Application The analysis informs the debate concerning how safe is safe enough for automated vehicles and provides initial indicators of what makes such vehicles feel safe and trusted.
... For example, workers may have to constantly overcome the opacity of algorithms and other systems that manage their work [2,27,45]. As such, gig labor platforms may be sources of tensions when users face information and power asymmetries [2,23]. These frictions are often the result of algorithms, which intentionally or unintentionally favor the interest of the platform over users [9,40,46]. ...
Full-text available
The algorithm-based management exercised by digital gig platforms has created information and power asymmetries, which may undermine the stability of gig work. Although the design of these platforms may foster unbalanced relationships, in this paper, we outline how freelancers and clients on the gig platform Upwork can leverage a network of alliances with external digital platforms to repossess their displaced agency within the gig economy. Building on 39 interviews with Upwork freelancers and clients, we found a dynamic ecosystem of digital platforms that facilitate gig work through and around the Upwork platform. We use actor-network theory to: 1) delineate Upwork's strategy to establish a comprehensive and isolated platform within the gig economy, 2) track human and nonhuman alliances that run counter to Upwork's system design and control mechanisms, and 3) capture the existence of a larger ecosystem of external digital platforms that undergird online freelancing. This work explicates the tensions that Upwork users face, and also illustrates the multiplicity of actors that create alliances to work with, through, around, and against the platform's algorithmic management.
Commuter comfort in cab rides affects driver rating as well as the reputation of ride-hailing firms like Uber/Lyft. Existing research has revealed that commuter comfort not only varies at a personalized level but also is perceived differently on different trips for the same commuter. Furthermore, there are several factors, including driving behavior and driving environment, affecting the perception of comfort. Automatically extracting the perceived comfort level of a commuter due to the impact of the driving behavior is crucial for a timely feedback to the drivers, which can help them to meet the commuter’s satisfaction. In light of this, we surveyed around 200 commuters who usually take such cab rides and obtained a set of features that impact comfort during cab rides. Following this, we develop a system Ridergo which collects smartphone sensor data from a commuter, extracts the spatial time series feature from the data, and then computes the level of commuter comfort on a five-point scale with respect to the driving. Ridergo uses a Hierarchical Temporal Memory model-based approach to observe anomalies in the feature distribution and then trains a Multi-task learning-based neural network model to obtain the comfort level of the commuter at a personalized level. The model also intelligently queries the commuter to add new data points to the available dataset and, in turn, improve itself over periodic training. Evaluation of Ridergo on 30 participants shows that the system could provide efficient comfort score with high accuracy when the driving impacts the perceived comfort.
Applications of the Digital Sharing Economy (DSE), such as Uber, Airbnb, and TaskRabbit, have become a main facilitator of economic growth and shared prosperity in modern-day societies. However, recent research has revealed that the participation of minority groups in DSE activities is often hindered by different forms of bias and discrimination. Evidence of such behavior has been documented across almost all domains of DSE, including ridesharing, lodging, and freelancing. However, little is known about the underlying design decisions of DSE platforms which allow certain demographics of the market to gain unfair advantage over others. To bridge this knowledge gap, in this paper, we systematically synthesize evidence from 58 interdisciplinary studies to identify the pervasive discrimination concerns affecting DSE platforms along with their triggering features and mitigation strategies. Our objective is to consolidate such interdisciplinary evidence from a software design point of view. Our results show that existing evidence is mainly geared towards documenting and mitigating issues of racism and sexism affecting platforms of ridesharing, lodging, and freelancing. Our review further shows that discrimination concerns in the DSE market are commonly enabled by features of user profiles and commonly impact reputation systems.
Conference Paper
Full-text available
Computational algorithms have recently emerged as the subject of fervent public and academic debates. What animates many of these debates is a perceived lack of clarity as to what algorithms actually are, what precisely they do, and which human-technology-relations their application may bring about. Therefore, this CSCW workshop critically discusses computational algorithms and the diverse ways in which humans relate to them—focusing particularly upon work practices and investigating how algorithms facilitate, regulate, and require human labor, as well as how humans make sense of and react to them. The purpose of this workshop is threefold: first, to chart the diversity of algorithmic technologies as well as their application, appropriation, use and presence in work practices; second, to probe analytic vocabularies that account for empirical diversity; third, to discuss implications for design that come out of our understandings of algorithms and the technologies through which they are enacted.
Full-text available
Unionization emerged as a way of protecting labor rights when society shifted from an agricultural ecosystem to one shaped by manufacturing and industrial labor. New networked work complicates the organizing mechanisms that are inherent to unionization. How then do we protect laborers from abuse, poor work conditions, and discrimination?
Conference Paper
Full-text available
This panel will explore algorithmic authority as it manifests and plays out across multiple domains. Algorithmic authority refers to the power of algorithms to manage human action and influence what information is accessible to users. Algorithms increasingly have the ability to affect everyday life, work practices, and economic systems through automated decision-making and interpretation of "big data". Cases of algorithmic authority include algorithmically curating news and social media feeds, evaluating job performance, matching dates, and hiring and firing employees. This panel will bring together researchers of quantified self, healthcare, digital labor, social media, and the sharing economy to deepen the emerging discourses on the ethics, politics, and economics of algorithmic authority in multiple domains.
Conference Paper
Full-text available
This paper contributes to the growing literature on peer-to-peer (P2P) applications through an ethnographic study of auto-rickshaw drivers in Bengaluru, India. We describe how the adoption of a P2P application, Ola, which connects passengers to rickshaws, changes drivers work practices. Ola is part of the 'peer services' phenomenon which enable new types of ad-hoc trade in labour, skills and goods. Auto-rickshaw drivers present an interesting case because prior to Ola few had used Smartphones or the Internet. Furthermore, as financially vulnerable workers in the informal sector, concerns about driver welfare become prominent. Whilst technologies may promise to improve livelihoods, they do not necessarily deliver [57]. We describe how Ola does little to change the uncertainty which characterizes an auto drivers' day. This leads us to consider how a more equitable and inclusive system might be designed.
Conference Paper
Full-text available
Apps allowing passengers to hail and pay for taxi service on their phone? such as Uber and Lyft-have affected the livelihood of thousands of workers worldwide. In this paper we draw on interviews with traditional taxi drivers, rideshare drivers and passengers in London and San Francisco to understand how "ride-sharing" transforms the taxi business. With Uber, the app not only manages the allocation of work, but is directly involved in "labour issues": changing the labour conditions of the work itself. We document how Uber driving demands new skills such as emotional labour, while increasing worker flexibility. We discuss how the design of new technology is also about creating new labour opportunities -- jobs -- and how we might think about our responsibilities in designing these labour relations.
Consumer-sourced rating systems are a dominant method of worker evaluation in platform-based work. These systems facilitate the semi-automated management of large, disaggregated workforces, and the rapid growth of service platforms-but may also represent a potential avenue for employment discrimination that negatively impacts members of legally protected groups. We analyze the Uber platform as a case study to explore how bias may creep into evaluations of drivers through consumer-sourced rating systems, and draw on social science research to demonstrate how such bias emerges in other types of rating and evaluation systems. While companies are legally prohibited from making employment decisions based on protected characteristics of workers, their reliance on potentially biased consumer ratings to make material determinations may nonetheless lead to a disparate impact in employment outcomes. We analyze the limitations of current civil rights law to address this issue, and outline a number of operational, legal, and design-based interventions that might assist in so doing.
Conference Paper
Online freelancing marketplaces have grown quickly in recent years. In theory, these sites offer workers the ability to earn money without the obligations and potential social biases associated with traditional employment frameworks. In this paper, we study whether two prominent online freelance marketplaces - TaskRabbit and Fiverr - are impacted by racial and gender bias. From these two platforms, we collect 13,500 worker profiles and gather information about workers' gender, race, customer reviews, ratings, and positions in search rankings. In both marketplaces, we find evidence of bias: we find that gender and race are significantly correlated with worker evaluations, which could harm the employment opportunities afforded to the workers. We hope that our study fuels more research on the presence and implications of discrimination in online environments.
The rise of the car-sharing company Uber will likely have mixed effects on labor standards. On the one hand, Uber’s partial consolidation of the car-hire sector and its compilation of data on passenger and driver behavior could enable the company and regulators to ensure safety and root out discrimination against passengers with relative ease. In that regard, Uber may be an improvement over the existing taxi sector, which is quite difficult to regulate. Uber’s longer-term impact on labor standards is quite unclear, however, and it may have dark implications for the future of low-wage work more generally.
Conference Paper
CSCW researchers have become interested in crowd work as a new form of collaborative engagement, that is, as a new way in which people’s actions are coordinated in order to achieve collective effects. We address this area but from a different perspective – that of the labor practices involved in taking crowd work as a form of work. Using empirical materials from a study of ride-sharing, we draw inspiration from studies of the immaterial forms of labor and alternate analyses of political economy that can cast a new light on the context of crowd labor that might matter for CSCW researchers.
Conference Paper
Mobile crowdsourcing markets (e.g., Gigwalk and TaskRabbit) offer crowdworkers tasks situated in the physical world (e.g., checking street signs, running household errands). The geographic nature of these tasks distinguishes these markets from online crowdsourcing markets and raises new, fundamental questions. We carried out a controlled study in the Chicago metropolitan area aimed at addressing two key questions: (1) What geographic factors influence whether a crowdworker will be willing to do a task? (2) What geographic factors influence how much compensation a crowdworker will demand in order to do a task? Quantitative modeling shows that travel distance to the location of the task and the socioeconomic status (SES) of the task area are important factors. Qualitative analysis enriches our modeling, with workers mentioning safety and difficulties getting to a location as key considerations. Our results suggest that low-SES areas are currently less able to take advantage of the benefits of mobile crowdsourcing markets. We discuss the implications of our study for these markets, as well as for "sharing economy" phenomena like UberX, which have many properties in common with mobile crowdsourcing markets.