lilly irani

lilly irani

intro’d to Lilly here:


New from @excinit @KernelMag: Can we build a humane alternative to Uber?… #platformcoop #IoO


News & info from the UCSD dept. of communication, plus conversations team-tweeted by two members of the community every 2 weeks | now: &

VC $$ backed platforms like Uber slash customer prices to gain market share. Hard for coops, upstarts to compete.…


.@ucsdCOMM @excinit @KernelMag time to anti-trust the platform monopolies and cooperativize the pieces. #platformcoop #IoO

and/or cooperativize us.. making the pieces irrelevant (pieces from article, ie: money, policy, et al)

are we meaning share as in share things or as in monetary shares.. huge difference. i’m not sure we’re ever sure which one we mean.


2014 – Stanford Seminar – Lilly Irani of UC San Diego

5 min – mechanical turk – way to make programmers managers – yet seemed exploitative.. so asked for workers bill of rights – one thing .. unfair/arbitrary rejections

9 min – programmer can hire 60 000 workers to do work in a couple days w/ mech turk.. mediation and arbitration fundatmentally unscalable.. promise of mt is large scale data labor for small operators and nimble actors

10 min – amt’s power – makes workers into a scalable computational infrastructure (launched by jeff bezo’s at mit)

11 min – image – a bunch of people on a chip – ads michael’s work with crowds… simply putting people on a chip and on a website.. doesn’t make that work.. when it came out people say.. oh like human api…

all power of coding at a machine.. now get people coding it

14 min – on having a lot of data to moderate.. hard for ai to do.. so you need large amounts of data labor

so amt is about making innovators.. org’s workers..  ie: workers as software stack as media

16 min – so more creative people are free to be creative.. while amt laborers do the (busy data work)

? whoa.

17 min – also mediates doubt – very powerful in making mech turk take off in really well meaning engineers…  invisible labor sustains the means of peer production in high tech

perhaps what amt represents… – because of piles of sub contractors

19 min – so how to proceed – w/workers on amt..

21 min – turkoptican – built (w/Michael) to interrupt invisibility – attempt to take this configuration that makes workers invisible.. and revisit… ie: w/reviewers

leaving a trail ness… via output via self-talk as data – but then too.. the people looking for people… much of that would become irrelevant…

24 min – on a more interpretive archive of what makes a worker good or bad


tool/web site more than just functional.. more than just way of evening the info playing field.. meant to use design as an intervention into the politics around crowdsourcing.. home page is deliberately agonistic  – rather than working toward consensus.. democracy works better if encourage strong pov’s

consensus ness

26 min – on workers from other countries.. mary gray

this is an intervention.. but ambivalent… not solution… changes conditions of reality..

27 min – other kinds of horizons/ecologies engaging in mutual aid w/in crowd work… one important place workers talk w/one another (listed forums) – organizing communities

28 min – emergent turker tactics…. forums make possible negotiation at a larger scale.. (ie: why didn’t do my job, why didn’t pay well)

what if these are – or could be – all irrelevant s

30 min – real lesson – valuing these systems – when we do these design interventions to improve people’s lives.. we tend to focus on software authors and credit them as heroes/entrepreneurs… bad for politics.. insulted workers… and possibility of moving forward in more collab way…

32 min – on crowd flash – to org online.. forums would tell me good idea.. some would do petition et al.. but usually peter out… so work on action plans.. motivate people to participate.. on the way..

or perhaps.. we come up with a nother way – where motivation becomes irrelevant..

33 min – conclusion:

  • making humans into computing infrastructure expands programmer agency while sustaining the myth of peer production
  • turkopticon interrupts worker invisibility and is one platform for mutual aid
  • worker run systems and emerging social computing reconfigure worker-employer relations

again – perhaps.. we come up with a nother way – where worker-employer relations becomes irrelevant..

37 min – q: how to design things into system so they don’t (on fairness of peer review) a: moocs.. uc learning at scale.. at end of day.. you can design some things into system.. ie: change econ incentives in the system… w/in that.. amt… social structure.. remains profitable to outsource work that is less valuable.. at some level econ structure is also a question here..

38  min – q: i’m not complaining about econ incentives.. i’m complaining about fairness… a: wages really are an issue… 2006 – fairness is what cut across… wages depended more on where you are living.. i don’t know if you can have human computation as a cheap platform w/o having these problems.. if you want to have an api served by humans and want it to be cheap…  so cheap fast or ethical.. you can only pick two i guess..

? – unless we look at deeper.. systemic change.. which all this tech is allowing now.. no?

ie: systemic change globally/equitably.. we focus on 3 elements:

qr problem deep

– – –

qr mechanism simple

– – –

qr system open

– – – –

be\cause… no?

42 min – if in an incentive structure… where porducing more.. scaling up is rewarded.. then relying on good will.. if someone could do that research… show that’s the case.. they could break through a political economy..

yeah. that.. what ness – we have enough research.. just need to be brave enough.. to try something different. for all of us.

43 min – hard to sample all workers… need very high access to employers.. a lot of methodological limitations.. turkopticon comes in from side.. how to change how real world system works w/o being in it

46 mi – q: if given your experience there are things you think we can do now to guide us more toward a utopian outcome… a: one of ways i look at this.. what kinds of people are becoming conventional… what do they afford.. vulnerabilities/risks do they produce… there are so many areas of research i’d do differently… ie: instead of thinking of human comp as new management science… i would ask .. what are forms of coop orgs – what would it look like to operate coop org .. and what kind of software needed for governance and distribution of payments… shifting focus to a just labor

or – what if a system where there are no payments.. no requestor id’s.. et al… io dance..  ps in the open.. mechanism simple enough..radical econ et al..

48 min – problem of way we think about where change and value comes from… looks into way we value intellectual property and patents.. how do we even frame innovation

property ness…

51 min – mech turk being more generalized skills.. anyone can do.. so need to oversee.


52 min – mech turk story i’m telling..  cutting down interactions to just parts that matter…

55 min – q: why do we have a limitation of what we see possible through crowdsourcing    a: i think .. part of pausisity of imagination… is the trade.. of coding… ie: why do crowdsource designers use crowdsourcing.. people come through limited pipeline…

API (application program interface) is a set of routines, protocols, and tools for building software applications. The API specifies how software components should interact and APIs are used when programming graphical user interface (GUI) components.

57 min – how to give crowd a sense of richness..

58 min – q: mech turk on max’ing pay… but a lot of workers do it for fun.. has there been any work on evaluating how fun tasks are… a: people have written or worker motivations.. how to make more tasks fun.. it’s not clear to me now that the majority of workers are doing it for fun.

yeah. a nother way.. doing whatever you want. that’s our energy\ness

today we can.  for (blank)’s sake

so we can’t not..

let’s do this firstfree art-ists.

(such an interesting experiment.. for all of us.. let’s try that.. short via short bp)

1:00 – q: look at atm.. and all the workers.. and subtract pay – how many would stay.. the answer is none

turkopticon is actually yelp for employers

1:03 – we’re not doing this research as thinkers in a pasture at a critical distance.. we’re responsible.. which risks are worth it and which aren’t… (ricardo dominguez) there are people in a more powerful position.. as someone sitting in academia.. i’d rather do the softer spoken thing…  key – when you talk about futures we design and futures we don’t… we need to be talking about these institutional accountability/risks that are shaping our research..  pretend it’s all about the user and it’s not


for goodness sake.. our children.. no?

gray research law et al



find/follow Lilly:

link twitter

Broadly, my research investigates the cultural politics of high-tech work practices with a focus on how actors produce “innovation” cultures. I work on these questions through two sites: entrepreneurial development efforts in India and the Amazon data processing outsourcing site Mechanical Turk. I am writing a book provisionally titled Entrepreneurial Citizenship: Innovators and their Others in Indian Development. I also sometimes collaboratively design, build, and maintain software (Turkopticon, Dynamo) as a means of understanding digitally mediated forms of work and their relationship to technological forms.