Tech Workers' Resistance Guide

$100,000 o he obo ge s i ! Tech Wo ke s’ Resis ance Guide:
Tech Wo ke Ac ions, His o y, Risks, Impac s, and he Case o a Radical Flank
Mohamed Abdalla
Uni e si y o Albe a
[email p o ec ed]
Abs ac
O e he pas decade, Big Tech has aced inc easing le els o
wo ke ac i ism. While wo ke ac ions ha e esul ed in pos-
i i e ou comes (e.g., cancella ion o Google’s P ojec D ag-
on ly), such successes ha e become inc easingly in equen .
This is, in pa , because co po a ions ha e adjus ed hei
s a egies o dealing wi h inc eased wo ke ac i ism (e.g., in-
c eased e alia ion agains wo ke s, and con ac s clauses ha
p e en cancella ion due o wo ke p essu e). This change
in company s a egy p omp s u gen ques ions abou upda -
ing wo ke s a egies o in luencing co po a e beha io in
an indus y wi h as socie al impac . Cu en discou se on
ech wo ke ac i ism o en lacks empi ical g ounding ega d-
ing i s scope, his o y, and s a egic calculus. Ou wo k seeks
o b idge his gap by i s ly conduc ing a sys ema ic analy-
sis o wo ke ac ions a Google and Mic oso epo ed in
U.S. newspape s o delinea e hei cha ac e is ics. We hen
si ua e hese ac ions wi hin he long his o y o labou mo e-
men s and demons a e ha , despi e pe cep ions o adical-
ism, con empo a y ech ac i ism is compa a i ely mode a e.
Finally, we engage di ec ly wi h cu en and o me ech ac-
i is s o p o ide a no el ca alogue o po en ial wo ke ac-
ions, e alua ing hei pe cei ed isks, impac s, and e ec i e-
ness (concu en ly publishing “Tech Wo ke s’ Guide o Re-
sis ance”1). Ou indings highligh conside able a ia ion in
s a egic hinking among ac i is s hemsel es. We conclude
by a guing ha he es ablishmen o a adical lank could in-
c ease he e ec i eness o cu en mo emen s.
In oduc ion
Tech Wo ke s a la ge U.S. echnology co po a ions (i.e.,
Big Tech)2a e o en po ayed as soli a y, highly indi id-
ualis ic, and apoli ical (Laiq 2024; And ews and Upadhya
2019; Khan 2018; Fung 2014). While low unioniza ion a es
(DiFelician onio 2021) ela i e o o he U.S. sec o s and a
1“Tech Wo ke s’ Guide o Resis ance” can be ound a h ps:
//www.cs. o on o.edu/~msa/TechWo ke sResis anceGuide.pd o
h ps://doi.o g/10.5281/zenodo.16779082
2In his wo k, we use collapse Selling and S imling’s ypology
o wo ke s: we conside all employees, con ac o s, and leade s as
ech wo ke s hough many o ou analyses ocus on pe sons wi h
high le els o echnical alen . We limi ou s udy o hose who
wo k in ‘Tech Fi ms’, pa icula ly Alphabe (hence o h: Google)
and Mic oso .
2000 2005 2010 2015 2020 2025
Yea
0
5
10
15
20
25
Numbe o Repo ed Wo ke Ac ions
Wo ke Ac ions by Yea
Google (P oQues )
Google (CAIT)
Mic oso (P oQues )
Mic oso (CAIT)
Figu e 1: Line plo o he numbe o ac ions by Google and
Mic oso wo ke s epo ed by majo U.S. Daily Newspa-
pe s (as cap u ed by P oQues ) pe yea om he la e 1980s
o Ma ch 22nd, 2025 and by Collec i e Ac ion in Tech (Tan
and Nedzh e skaya 2020) om he la e 1980s un il 2023.
his o ical eliance on indi idual sca ci y o e collec i e sol-
ida i y (Doc o ow 2025) ha e long de ined he ech sec-
o , including AI esea che s (Tie nan 2019), ecen head-
lines umpe a shi , showcasing employees mobilizing o
causes hey champion (Bomey 2019; Wakabayashi e al.
2018; Scheibe and Conge 2020).
The snowballing hype a ound AI, massi e inancial in-
es men s, and in ensi ying compe i ion be ween majo
echnology companies is p ojec ed o d ama ically ampli y
AI’s socie al impac , he eby ele a ing he impo ance o
wo ke ac i ism. Un o una ely, compa ed o ea lie in-
s ances o wo ke ac i ism which we e success ul in quickly
achie ing hei goals (e.g., cancella ion o Google’s P ojec
D agon ly and non- enewal o P ojec Ma en) i appea s
ha cu en e o s ha e been less e ec i e. This is due,
in pa , o co po a ions adjus ing hei s a egies o dealing
wi h inc eased ac i ism (e.g., e alia ing agains ac i is s and
pu ing in con ac clauses ha p e en e mina ion based
on wo ke p essu e). This change p omp s c i ical ques-
ions abou he u u e o wo ke ac i ism (bo h i s na u e
and scope). Wha o m o ac ions a e wo ke s aking and
how should his change? How do he s a egies employed by
wo ke s oday compa e o hose employed by his o ical ac-
ions? Wha can wo ke s oday lea n om hose in he pas ?
Indeed, compa ed o he con on a ional ac ics employed
by pas labou and social mo emen s3, cu en ech ac i is s
appea hesi an o emb ace ac ion ha isk being labelled
iolen , des uc i e, o o e ly dis up i e.
This wo k aims o enhance he e ec i eness o wo ke ac-
ions. We belie e his can be achie ed by: a) compa ing how
he equency, o ms, and e olu ion o con empo a y wo ke
ac ions compa e agains a his o ical in en o y o ac i is ac-
ics, and b) engaging wi h cu en and o me ech ac i is s
o analyze he pe cei ed isks, impac s, and e ec i eness o
a ious ac ions, he eby in o ming s a egic decision making
and educa ion.
Compa ing his o ical and cu en app oaches e eals ha
while ech wo ke ac ions a e inc easing in hei numbe ,
hei b ead h emains na ow. We also obse e ha while
he e is subs an ial a ia ion in pe cep ion o isk and im-
pac o ac ions, he e appea s o be appe i e o mo e adical
ac ions. Combining hese insigh s, his wo k concludes by
a guing ha he eme gence o a “ adical lank” could sub-
s an ially inc ease he e ec i eness o u u e ech indus y
ac i ism.
De ining Wo ke Ac ion
Pas wo k explo ing he ac i ism o ech wo ke s and AI e-
sea che s, has ei he ocused on a hand ul o selec ed ac-
ions (Bel ield 2020), o a speci ic ype o ac ion (e.g., col-
lec i e ac ion (Boag e al. 2022)). In con as , ou wo k akes
a b oade lens o wha quali ies as an ‘ac ion’. We de ine po-
li ical ac ion (hence o h sho ened o ac ion o ac ) as any
pu pose ul ac ion aken by a pe son o g oup o people o
ad ance a poli ical o social goal. Some examples o ech
wo ke ac ions include:
• Jack Poulson, o me senio Google scien is , qui ing
o e “ o ei u e o ou alues” ega ding an e o o
launch a censo ed sea ch engine in China (Gallaghe
2018).
• Ke in Ce nekee, o me Google enginee , speaking o he
p ess accusing he company o i ing him o exp essing
conse a i e poli ical belie s (Gha a y 2019).
• F ances Haugen, o me Facebook da a scien is , whis le-
blowing o e eal ha Facebook’s own esea ch showed
ha i ampli ies ha e, misin o ma ion and poli ical un es .
(Pelley 2021; Olesen 2025).
• Google employees s aging si -ins a company o ices o
p o es cloud con ac s wi h is ael’s go e nmen (Haskins
2024).
• Joe Lopez, o me Mic oso enginee , in e up s CEO’s
keyno e wi h p o-Pales inian p o es (Bhuiyan 2025).
As can be obse ed, he e is a ia ion in he numbe o
ac o s (one s many), o o ms ( om si -ins o leaks), and o
pu pose – no single poli ical s ance can ie hem all oge he
( hough U.S. ech wo ke s ha e been shown o lean owa ds
he Democ a ic pa y in he U.S. (Fe ens ein 2015)).
3The i le o his piece in homage o he ac ion aken by NYU
s uden s who occupied a se e oom and held a ansom o he
Black Pan he s asking o “$100,000 o he obo ge s i !” (Julyk
2008; Ba on 2015)
Cu a ing His o ical Ac ions
To cu a e a lis o his o ical ac ions, we elied on wo ap-
p oaches. Fi s , we elied on a s udy o ele an books
(Malm 2021; Collinson and Ack oyd 2005; Ro hs ein 2022;
Ha is 2023), a icles/w i ings (CIA 1944; Sha p and Finkel-
s ein 1973), and academic publica ions (Boag e al. 2022;
Bel ield 2020). All sou ces used o his sec ion we e ead
wi h he in en ion o ansla ing ac ions om hei his o ical
con ex o he cu en ech se ing.
To complemen he analysis o li e a u e, we sough ou
expe inpu . Speci ically, we engaged cu en and o me
employees o he Big Tech co po a ions (n=8) o suppo
he iden i ica ion and e alua ion o po en ial wo ke ac-
ions. This s udy was app o ed by he Uni e si y o Al-
be a REB (#P o00151408). In con as o Sha p and Finkel-
s ein (1973)’s de ailed lis o indi idual ac ions, we op ed o
b oad ca ego ies (e.g., classi ying bo h dis up i e wo kplace
si -ins and he occupa ion o a VP’s o ice as ‘Dis up i e
Wo kplace P o es s’). We chose his b oade ca ego iza ion
o h ee p ima y easons: (a) he c ea ion o a uly com-
p ehensi e lis o all speci ic wo ke ac ions p esen s a nea -
insu moun able challenge; (b) such a lis would be oo la ge
o e ec i ely analyze; and (c) we an icipa e ha a highly
g anula lis would no e ain i s u ili y o u u e wo ke s, a
cen al aim o his s udy. Table 1 p esen s a summa y o he
possible ypes o ac ions ha can be aken by ech wo ke s.
Analyzing His o ical Ac ions Fo each wo ke ac ion, we
elied on expe judgmen o desc ibe he ac ion as posi i e
o nega i e4, and explica e he possible isks, impac , and
o e all e ec i eness o he ac ion. This p ocess ensu ed ha
he analysis was g ounded in bo h heo e ical unde s anding
and p ac ical, expe -based knowledge. The analysis o each
ac ion can be ound in he “Tech Wo ke s’ Guide o Resis-
ance”.
We g ouped ac ions in o highe le el g oups based on
commonali ies be ween hem (e.g., all a e some so o
p o es , e c.). These g oupings a e no p esc ip i e, bu a he
a way o o ganize he da a o make i easie bo h o sea ch o
ac ions and mo e quickly unde s and he a ie y a ailable o
wo ke s.
S udying Ac ions Taken by Tech Wo ke s
Despi e he s e eo ype o ech wo ke s being la gely apoli -
ical, he e is his o ical p ecedence o ech/scien i ic wo ke s
o en pa aking in poli ical ac ions. Fo example, i was a
pai o Pola oid employees – o which one was echnical in
na u e (a chemis ) – who launched he i s an i-apa heid
boyco o a U.S. co po a ion (McCanne 2020). Likewise,
in he la e 1990s, Mic oso los a class-ac ion sui o in-
co ec ly classi ying employees as con ac o s (McCoy and
BankS a 1997). Thus, ou i s cou se o ac ion is o i mly
es ablish he ac i i y le el o ech wo ke s, and he ypes
4Nega i e ac ions seek o hal o dis up he no mal ope a ions
o a company by e usal o obey o de s o obs uc ing o de s. Posi-
i e ac ions seek o build al e na i es o exis ing sys ems o epai -
ing ela ionships. This is discussed in mo e de ail in he Discussion.
Qui ing Law a e
Qui (quie ly) (I i ed) Sue
Qui (in e nal p o es ) Lawsui s
Qui (ex e nal p o es ) Repo ing o Regula o y Body
Engage wi h HR
Lobbying
P o es ing Malicious Compliance
Public P o es s Minimal p oduc i i y (Wo k- o- ule)
Public P o essional P o es s Collec i e delays
In e nal P o es s (ac i e) Hi ing lowe quali y candida es
In e nal P o es s (passi e) Push o qui ing
Dis up i e Wo kplace P o es s Elonga e bu eauc a ic p ocesses
Bai ing Reac ions
S iking
Knowledge T ans e Violence
Leaking Anonymously Towa ds In as uc u e
Leaking Publicly Towa ds Sel
Publicize Ac ions Towa ds O he s
Educa ional campaigns/ each-ins Sabo age p oduc s
Sabo age in e nal sys ems
Building Communi y Posi i e Ac ions
O ganize Build Compe ing P oduc s
Engage wi h ac i is /ci il igh s g oups C ea e assis i e ools/ esou ces
Hi ing Poli ically Aligned S a Repe a i e Ac ions
Table 1: A summa y o possible ypes o ac ions ha can be aken by ech wo ke s.
o ac ions unde aken. As men ioned p e iously, ou explo-
a ion di e s in 3 signi ican ways om p io wo k: a) we
seek o be expansi e in ou co e age o he ac ions aken by
ech wo ke s, and b) we do no limi ou analysis solely o
collec i e ac ions, no do we c) limi ou analysis o non-
iolen ac ions.
Da a and Me hods
To ensu e a s anda dized and ep oducible app oach, we de-
cided o le e age media co e age o ac ions a wo i ms
(Google and Mic oso ) o wo easons: Fi s , hese i ms
a e amongs he mos well-known and comp ehensi ely-
co e ed US ech companies in news media. Second, since
we a e manual il e ing h ough a icles, p ac ical necessi y
compels us o limi he numbe o companies, accep ing he
educed gene alizabili y o such an analysis.
Fo his explo a ion, we elied on P oQues ’s U.S.
Newss eam (P oQues Unda ed). The U.S. newss eam is
an a chi e o U.S. news con en s e ching back om he
1980s o p esen -day and con ains ull- ex news sou ces
o well-known na ional newspape s such as The New Yo k
Times and The Wall S ee Jou nal, as well as o e 80 e-
gional and local news sou ces. We used he ollowing que y:
(google OR alphabe OR mic oso ) AND (wo ke OR em-
ployee OR con ac o OR s a OR g oup). We limi ed ou
esul s’ sou ces o ull- ex newspape s which we e w i en
in English and discussed he subjec o ‘employees’ as de-
ined by he P oQues ‘Subjec ’ il e . We conduc ed he ini-
ial sea ch on Ma ch 22nd o 2025 and conside ed all esul s
published be o e his da e.
Ou que y esul ed in 11,859 news a icles. We i s used
i les o iden i y ele an a icles (i.e., hose ha discussed
wo ke ac ions). Fo a icles iden i ied as possibly ele an
(n=1,201), we ead he ull a icle o de e mine i hey we e
uly ele an . Fo his limi ed subse (n=154), we ex ac ed
all men ioned wo ke ac ions and associa ed da es. This p o-
cess was conduc ed by wo people and discussions we e
used o esol e disag eemen s.
We chose o cap u e ou da a om news co e age in-
s ead o using he exis ing Collec i e Ac ion in Tech (CAIT)
da ase (Tan and Nedzh e skaya 2020), a da abase which
documen ed collec i e ac ions om wo ke s in he ech in-
dus y, because ou scope is wide . CAIT ocuses solely on
collec i e ac ions which is na owe han he scope o ou
pape : he e a e wo ke ac ions (e.g., qui ing o whis le-
blowing) ha o en a e no aken in a(n explici ly) collec i e
manne . An unin ended bene i o his choice is ha we a e
also able o explo e he co e age o collec i e ech ac ions
by he collec ion o majo U.S. newspape s cu a ed by P o-
Ques (as opposed o he sea ch execu ed by Nedzh e skaya
and Tan (2021) using NexisUni).
Resul s
Ac ions o e ime Figu e 1 plo s he numbe o ac ions
aken by wo ke s a Google and Mic oso o e he pas 40
yea s ha ha e been epo ed by la ge U.S. daily newspa-
pe s. We can obse e ha he la e 2010s ep esen he peak
o epo ed wo ke ac ions. We can also see ha Google em-
ployees end o pa ake in mo e ac ions ela i e o Mic oso
employees. Bo h Google and Mic oso wo ke s we e mos
ac i e (as measu ed by co e age in news epo s) 2015 and
onwa ds, likely indica ing a cul u al shi o so s o a change
2000 2005 2010 2015 2020 2025
Yea
Lawsui
Leaking
Public
P o es
In e nal
P o es
Passi e
Qui
Ac ion Type
(a)
2000 2005 2010 2015 2020 2025
Yea
Wo ke
En i omen
Di e si y
Sexual
Ha assmen
Mili a y
Pales ine
Ac ion Topic
(b)
Figu e 2: His og am o he numbe o ac ions by (a) ype and (b) ocus/mo i a ion as epo ed by majo U.S. daily newspape
(as cap u ed by P oQues ) pe yea om he la e 1980s o Ma ch 22nd o 2025.
in he wo k o hese companies. Unco e ing he speci ic ea-
son o his shi is impo an wo k ha is beyond he scope
o his pape . A bo h companies, we can see ha a e he
la e-up p e-2020, he e has been a ela i e down- u n in he
numbe o wo ke ac ions. This could be caused by he in-
c eased c ackdown on p o es s o o ce ain speech (Mohan
2024; Saina o 2024; Meh a 2025), o newspape s’ choice o
no epo on hese ac ions. This highligh s an impo an lim-
i a ion o his app oach (edi o ial choice is a con ounding
ac o which we a e unable o co ec o ).
Types o ac ions Tech wo ke s employed a a ie y o
s a egies5 o achie e hei goals. O he 153 ac ions
obse ed in ou da ase , 22% (n=33) in ol ed whis le-
blowing/leaking o he media anonymously and an equal
pe cen age, 22% (n=33), we e ‘In e nal Passi e P o es s’
(e.g., pe i ions, discussions on in e nal o ums, e c.). This
was ollowed by aking he employe o cou a 16% (n=25).
We can see ha he b ead h o ac ions aken by ech wo k-
e s has inc eased, Figu e 2(a). Be o e 2015, he mos com-
mon ac ions we e ‘Lawsui ’ and ‘Leaking’, whe eas, a e
2015, he mos common ac ions we e ‘Leaking’ and ‘In e -
nal P o es (Passi e)’ wi h he in oduc ion o public p o es s
and qui ing. The ull able can be ound in Appendix Table
1.
Ac ion Causes By a he mos common cause is ‘Wo ke
En i onmen ’ wi h nea ly a hi d (28%, n=43) o ac ions. We
de ine ‘Wo ke En i onmen ’ o include conce ns su ound-
ing pay, bene i s, and employmen s a us. This was ollowed
by 16% (n=24) o ac ions ela ing o di e si y o iden i y-
based issues (e.g., ace, gende , o age). P o-Pales ine and
an i-wa ac ions (i combined) ake he hi d mos common
p o es cause wi h 14% (n=22) o all ac ions – wi h p o-
Pales ine ac ions occu ing a mo e han double he a e o
gene al an i-wa ac ions. ‘Sexual Ha assmen ’ has also been
5No e: o his sec ion we a e using he amewo k de eloped
in he p e ious sec ion o he he pape .
a la ge sou ce (10%, n=15) o ac ions. Like he ypes o ac-
ions, he mo i a ion o ac ions has also changed o e ime.
While desi e o imp o emen s in he wo k en i onmen and
di e si y consis en ly mo i a e wo ke ac ion, ecen social
changes and wide wo ld e en s ha e p ecipi a ed wo ke
ac ions, as seen in he ac ions ha eme ged ollowing he
peak o he #me oo mo emen (2017-2018), he mu de o
Geo ge Floyd (2020), and is aeli a acks on Pales ine (2021,
2023–P esen day (2025)). The ull able can be ound in
Appendix Table 2.
Quan i ying Pe cei ed Risk and Impac
Ha ing colla ed a b oad se o ac ions, ou aim was o asce -
ain expe pe spec i es on hei po en ial impac and he ac-
companying isk. Is he e wide ag eemen on impac o isk?
Does high-impac necessa ily en ail high isk? Does he u -
gency o a cause a ec he pe cei ed impac o ac ions? Fo
he sake o b e i y in his sec ion, we will ocus solely on
gene al impac ; howe e , he same p ocedu e was employed
o quan i y pe cei ed isk and pe cei ed immedia e impac .
Ins ead o using a adi ional Like scale e alua ion
me hodology asking expe s o choose be ween a high/medi-
um/low scale (which would equi e us o p e-de ine each
ca ego y, hus possibly biasing he esponden s), we in-
s ead used compa a i e anno a ions (Thu s one 1927; Da id
1963). In i s simples o m, compa a i e anno a ions can be
buil as pai ed compa isons which p esen s anno a o s wi h
wo op ions and asks hem o choose which i em is g ea e
wi h espec o he p ope y o in e es (e.g., impac ). Wi h
mul iple ques ions, we would be able o gene a e an o dinal
anking o i ems. In his wo k, we use Bes -Wo s Scaling
(Lou ie e and Woodwo h 1991), an ad anced o m o com-
pa a i e anno a ions ha expands he wo-i em compa a i e
anno a ion ask o mul iple i ems. In his wo k, anno a o s
a e gi en n=4 i ems (i.e., ac ions) and asked o choose which
ac ion is mos and leas impac ul. Each anno a ion o a 4-
uple p o ides us wi h i e pai wise inequali ies. Fo exam-
Figu e 3: Pe cei ed gene al impac and isks o he ull se o possible ac ions.
ple, i among a, b, c, and d,ais selec ed as he mos impac -
ul and dis selec ed as leas impac ul, hen we know ha
a > b, a > c, a > d, b > d, and c>d. F om his, we can
calcula e an o dinal anking o ac ions using a simple coun -
ing mechanism: he ac ion o imes an ac ion was chosen
as mos impac ul minus he ac ion o imes he i em was
chosen as leas impac ul (O me 2009). P e ious wo k has
shown ha eliable sco es a e ob ainable om abou 2N4-
uples (Ki i chenko and Mohammad 2016, 2017).
F om ou ull lis o 37 ac ions, we gene a ed 74 unique
4- uples6(each 4- uple consis ing o 4 dis inc ac ions), wi h
each ac ion appea ing in abou 8 o he 4- uples. Anno a o s
we e no p o ided de ailed o echnical de ini ions o impac
o isk; a he , we encou aged anno a o s ely on hei ex-
pe ise and indi idual pe spec i es. All anno a o s o his
ask a e all cu en o ecen ly- o me employees a Google
and Mic oso who ha e pa icipa ed in ac ions (ei he indi-
idually o as a collec i e). 5 anno a o s comple ed labeling
he ei he se o 74 unique 4- uples, wi h he emainde only
anno a ing a small pe cen age.
Resul s
Reliabili y o Anno a ions
Since ou compa a i e anno a ion p ocedu e gene a es eal-
alued sco es, we used spli -hal eliabili y (SHR) o mea-
su e he quali y o quali y and eliabili y o ou anno a ions
6The uples we e gene a ed using he BWS sc ip s p o ided by
Ki i chenko and Mohammad (2017): h p://sai mohammad.com/
WebPages/Bes Wo s .h ml. This esou ce was also used o calcu-
la e he spli -hal eliabili y and he o dinal anking o ac ions.
(C onbach 1951; Kude and Richa dson 1937). SHR mea-
su es he deg ee o which epea ed anno a ions o he same
ask would esul in simila ela i e anking o i ems. To
measu e SHR, o each 4- uple, we spli he 4- uples in o
wo bins. Each bin was used o independen ly p oduce he
eal- alued sco es o each i em which we e compa ed us-
ing Spea man co ela ion. This p ocess was epea ed 1000
imes.
Ou anno a ions had e y high SHR: Spea man co ela-
ion (SpC) o 0.92 (i.e., ‘ e y s ong’) o bo h ankings o
pe cei ed isk and impac . The high eliabili y o anno a-
ions, despi e he lack o explici ins uc ion, indica es ha
he e is a sha ed iew in he pe cei ed isk, impac , and im-
media e impac o ac ics. No e: despi e he gene ally high
le el o ag eemen on he anking o ac ions, he e we e
some s ong indi idual di e ences o ce ain ac ion (e.g.,
Violence Agains Sel ’s impac anging om -1 o 0.25), Fig-
u e 4.
Risk s Impac
Figu e 3 plo s, on a wo dimensional g id, he pe cei ed
isk agains he pe cei ed impac o ac ions. Appendix Table
3 p esen s he quan i ied sco es o pe cei ed isk, impac ,
and immedia e impac o each ac ion. The e is gene ally a
s ong co ela ion be ween he pe cei ed isk and he pe -
cei ed impac o an ac ion (SpC o 0.58) which inc eases
u he o isk s immedia e impac (SpC o 0.65). Figu e 4
plo s he pe cei ed isk agains he pe cei ed impac o ac-
ions (la ge do s) spli by ac ion ca ego ies. Fo he i e an-
no a o s who comple ed he ull se o 74 anno a ions (370
bina y compa isons), we place hei pe sonal anking o ac-

1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
(a)
Building Communi y
Ac ion
Unionize
Engage wi h ac i is /ci il igh s g oups
Hi ing Poli ically Aligned S a
1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
(b)
Knowledge T ans e
Ac ion
Leaking Anonymously
Leaking Publicly
Publicize Ac ions
Educa ional campaigns/ each-ins
1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
(c)
Law a e
Ac ion
(I i ed) Sue
Lawsui s
Repo ing o Regula o y Body
Engage wi h HR
Lobbying
1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
(d)
Malicious Compliance
Ac ion
Minimal p oduc i i y
Hi ing lowe quali y candida es
Push o qui ing
Elonga e bu eauc a ic p ocesses
Boyco (nega i e)
1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
(e)
Posi i e Ac ions
Ac ion
Build Compe ing P oduc s
C ea e assis i e ools/ esou ces
Repe a i e Ac ions
Boyco (posi i e)
1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
( )
P o es ing
Ac ion
Public P o es s
Public P o essional P o es s
In e nal P o es s (ac i e)
In e nal P o es s (passi e)
Dis up i e Wo kplace P o es s
Bai ing Reac ions
S iking
Re use o build
1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
(g)
Qui
Ac ion
Qui (quie ly)
Qui (in e nal p o es )
Qui (ex e nal p o es )
1.00 0.75 0.50 0.25 0.00 0.25 0.50 0.75 1.00
Risk
1.00
0.75
0.50
0.25
0.00
0.25
0.50
0.75
1.00
Impac
(h)
Violence
Ac ion
Violence Towa ds In as uc u e
Violence Towa ds Sel
Violence Towa ds O he s
Sabo age p oduc s
Sabo age in e nal sys ems
Figu e 4: Pe cei ed gene al impac and isks o he ull se o his o ical ac ions spli by ac ion ca ego y. Wi hin each ca ego y,
he la ge do s ep esen he o e all anno a ion (i.e., conside ing all anno a o s) o each indi idual ac ion. The smalle shapes
ep esen an indi iduals a ing o an indi idual ac ion. Each anno a o is gi en a consis en shape, and each ac ion is gi en a
consis en colo . We used a KDE (Ke nel Densi y Es ima e) plo o isualize he p obabili y densi y unc ion o isk and impac
( he blue lines). A la ge e sion o his igu e can be ound in Appendix Figu e 1.
ions esul ing om hei labels alone using smalle ma ke s
(e.g., small do s, x’s, squa es, e c.).
Building Communi y This amily o ac ions exhibi s a
posi i e co ela ion be ween pe cei ed isk and impac (SpC
o 0.5). Gene ally, his amily o ac ion was iewed as ha -
ing minimal isk ( hough unioniza ion was conside ed high
isk). Building communi y was a ed as he second mos im-
pac ul ac ion gene ally, howe e , when seeking u gen im-
pac , i d opped o he middle o he pack.
Knowledge T ans e This amily o ac ions exhibi s a
mode a e co ela ion be ween pe cei ed isk and impac
(SpC o 0.40). As a amily o ac ions, i is among he mos
impac ul e en when u gen impac is needed, while ha -
ing median isk. Wi hin his amily o ac ions, leaking (i.e.,
whis le-blowing) had high impac , and he impac (and isk)
is inc eased u he when done publicly.
Law a e Law a e did no exhibi a s ong co ela ion be-
ween pe cei ed isk and impac (SpC o 0.20). As a amily
o ac ions, law a e was amongs he lowes in e ms o im-
pac (d opping o lowes o immedia e impac ), while being
nea he median o isk o he wo ke . Wi hin his amily,
lobbying was pe cei ed o ha e he highes posi i e impac .
Malicious Compliance As a amily o ac ions, he co e-
la ion be ween pe cei ed impac and isk was weak (SpC o
0.30). Ac ions in his amily a e pe cei ed o be he lowes
impac (e en o immedia e impac ) and nea lowes isk.
Posi i e Ac ions Posi i e ac ions had a posi i e co ela-
ion be ween pe cei ed isk and impac (SpC o 1.0). Ac-
ions wi hin his amily we e deemed o ha e he lowes
isk and medium impac (including o immedia e impac ).
“Building compe ing p oduc s” s ood ou om he his am-
ily o ac ions as ha ing high impac and ela i ely low isk.
P o es P o es ing as a ca ego y o wo ke ac ion has a
e y posi i e co ela ion be ween pe cei ed isk and impac
(SpC o 0.88). As a amily, p o es ing was deemed o be
somewha impac ul, ising o he second mos impac ul se
o ac ions when immedia e impac was needed. This amily
was pe cei ed o be hi d iskies . The e is howe e , g ea
a ia ion (wi h espec o pe cei ed isk and impac ) wi hin
his amily. P o es s which a e dis up i e (e.g., ‘S iking’ o
‘Dis up i e Wo kplace P o es ’) we e bo h he highes im-
pac and he highes isk. The impac and isk o ac ions co -
ela ed wi h how ied o he message an ac ion could be and
how much i a ec ed he wo kplace.
Qui Qui ing had a s ong co ela ion be ween pe cei ed
impac and isk (SpC o 1). Qui ing was pe cei ed o be he
second iskies amily o ac ions, while being anked in he
lowe hal o impac (nea median o immedia e impac ).
The e is a la ge ange in impac despi e a much na owe
a ia ion in isk. Qui ing as a o m o in e nal o ex e nal
p o es was seen o be much mo e e ec i e han qui ing qui-
e ly despi e only a small change in isk.
Violence This amily o ac ions exhibi ed only a weak co -
ela ion be ween pe cei ed impac and pe cei ed isk (SpC
o 0.09). This ca ego y o ac ions was bo h he mos impac -
ul ( o bo h long and immedia e impac ) and he highes
isk. “Violence agains onesel ” was a ed as he leas e ec-
i e ac ion o his amily.
Discussions – Mo als, S a egy, Risks
Thus a , we did no expound on wha wo ke s should do.
Compa ing wha has been done (§3) agains wha could done
(§4) highligh s he oom o expanding he b ead h o ac-
ions. The quan i ica ion o pe cei ed isk s impac (§5)
can be used o de e mine whe e one’s ole ance lies, and
wha ac ions a g oup can ake wi hou aliena ing i s mem-
be s. Un o una ely, s a egic hinking is highly con ex ual;
wo ke s exis in di e en con ex s, ace di e en s esso s
and ha e di e en goals. Fo example, wo ke s on isas ace
di e en isks han hose who a e ci izens ( ega ding depo -
a ion o hei ac i ism as demons a ed by he ecen depo -
a ions o p o-Pales ine ac i is s (AlJazee a 2025))7. Like-
wise, employees who a e esponsible o dependen s a e op-
e a ing unde di e en ci cums ances han hose who don’
ha e such a esponsibili y.
The e o e, in his sec ion, o enable wo ke s o assess
o hemsel es how o p io i ize ac ions, assess e ec i e-
ness, and de elop cohe en s a egies, we discuss se e al
commonly aised opics and ques ions ela ed o wo ke ac-
i ism. I is c ucial o no e ha his discussion and ad ice
is pu pose ully kep a a high-le el. Indi iduals mus ac-
coun o hei speci ic con ex s, isk ole ances, and legal
ju isdic ions.
Mo al Inju y and The Need o Ac
Mo al inju y e e s o when “ he e has been (a) a be ayal
o ‘wha ’s igh ’; and (b) ei he by a pe son in legi ima e au-
ho i y [...], o by one’s sel , (c) in a high s akes si ua ion”
(Shay 2014; Li z e al. 2009). Mo al inju y has been expe i-
enced by ech wo ke s when hey lea n ha he co po a ions
o whom hey wo k ad ance causes hey a e mo ally agains
(e.g., con ibu ing o clima e change (S one 2024), o a ming
is ael (Ib aheem 2025; Wa en and Ro h 2025)).
Sel - e lec ing on his own e hical decisions as a com-
pu e science g adua e s uden , Chan (2020) makes he dis-
inc ion be ween mo al obliga ions and mo al aspi a ions.
Mo al obliga ions a e ac ions whe eby an indi idual be-
comes mo ally blamewo hy should hey ail o pe o m
hem (o in he case o a nega i e ac ion: he blame is as-
signed when an ac ion is pe o med). On he o he hand,
mo al aspi a ions e e o ac ions whe e an indi idual is
deemed mo ally i uous o doing hem, bu no mo ally
blamewo hy o no pe o ming hem. He no es, o exam-
ple, he has a mo al obliga ion o no wo k on ‘building AI
sys ems ha enable genocide’, bu only a mo al aspi a ion o
pu sue a esea ch di ec ion which maximizes ‘posi i e e hi-
cal impac ’ (Chan 2020). Th ough his examples, Chan high-
7While his has adi ionally held ue, i seems ha he go -
e ning U.S. adminis a ion is conside ing depo ing U.S. ci izens
(Rubin 2025).
ligh s he wide a ie y o seemingly mundane choices made
by g adua e s uden s ha en ail e hical consequences (e.g.,
selec ing esea ch di ec ions, seeking unding, and men o -
ing).
Inspi ed by an Islamic pa adigm, o me Google so wa e
enginee Hasan Ib aheem a i es a a simila conclusion. Re-
lec ing upon his wo k and his con ibu ions o Google’s “ di-
ec suppo o he genocide agains Pales inians due o hei
p io i iza ion o p o i o e human li es” (Ib aheem 2025),
he concludes ha despi e no di ec ly wo king on mili a y
echnology, p o iding his labou o a company complici
in immo al ac ions makes him complici . He sugges s ha
wo ke s should aim o qui o ge i ed in as loud a manne
as possible as, pe his piece, “[o] ganizing does no absol e
you o complici y inde ini ely” (Ib aheem 2025).
Fo hose who eel s ongly abou pa icula issues such as
Chan and Ib aheem, i ’s clea ha con ibu ing o such com-
panies, e en in angen ial ways, may s ill esul in expe ienc-
ing mo al inju y. When such inju y canno be a oided, mo al
inju y can be coun e -ac ed by aking ac ions which accoun
o mo al epai (Li z e al. 2009; Walke 2001). Mo al epai ,
a opic wi h ich li e a u e, can be (o e ly) simpli ied o do-
ing good deeds (mo e speci ically app op ia e amends) o e-
gain posi i e sel -judgmen abou one’s sel (Vi es-Gab iel,
Van Len , and We s ein 2023). Con inuing wi h Ib aheem’s
Islamic heme, his would be analogous o he Islamic con-
cep o awba (which can be incomple ely ansla ed o e-
pen ance) (Abdullah 2022) whe e, in addi ion o emo se
and abandonmen o he ac ion causing mo al inju y, he pe -
son should also seek o make amends o any ha ms hey
ha e caused (Tahi 2022). We belie e ha any o he ac ions
discussed in his wo k can help mo al epai bu his depends
on a a ie y o ac o s (e.g., belie ha he e is ne posi i e
impac , which may be di icul o see).
Gi en he na u e o his pape , we ha e p io i ized he
oices o wo ke s who ha e been in ol ed in wo ke ac-
ions o ele an discussions. This ocus ep esen s a sam-
pling bias o so s. I may be possible ha mos wo ke s do
no eel s ongly abou any pa icula issue and hus su e
no mo al inju y. In his case, o maximize impac o ac ions,
i would be bene icial o wo ke s o ins ill sha ed alues in o
colleagues such ha pa aking in he sys em esul s in mo al
inju y hus d i ing hem o ac (i.e., mo al educa ion).
High-Le el s a egy
While s udying he di e ences be ween cu en and his o -
ical ac ics is enligh ening, i does no necessa ily in o m
cu en and u u e ech ac i is s on how o engage. Below,
we s a wi h a b ie high-le el ision o how ac ions should
be app oached.
Fi s , ac i is s should e lec on wha d i es hem o ac .
A e hey d i en ou o emo ion? Guil ? Ego? While di e en
people may ha e di e en sou ce mo i a ions ensu ing cla -
i y in one’s mo i a ion can lead o sus ained ac i ism (Jin
e al. 2025; Hall 2019). Wha is he mo al o poli ical ame-
wo k h ough which hey see he wo ld? Di e en ame-
wo ks will p io i ize o disallow ce ain ac ions, and ha e a
g ea e ec in he balancing o p io i ies. Wha obliga ions
and es ic ions do you ha e placed upon you and how may
hey a ec which ac ions you may pa ake in?
Then, i ’s impo an o conside he exis ing incen i e
s uc u es o he sys ems you’ e ying o impac . The ma-
jo i y o humans and co po a ions a e no (consciously) mo-
i a ed by poli ical ambi ions o iews. Ra he , hey eac o
incen i es and dis-incen i es placed upon hem by he sys-
em in which hey li e and wo k. Unde s anding ha sys em
and he s uc u es ha d i e decision making will enable ac-
i is s o map ou he powe dynamics o hei socie y and
be e simula e, si ua e, and execu e o hei cause (Bolsen
2025).
Las ly, ac i is s should se hei plans a he s uc u al
le el. I ’s impo an o ealize ha all impo an causes e-
qui e “long e m, edious, us a ing, exaspe a ing, bo ing,
disappoin ing an i-clima ic wo k o e decades o achie e
inc emen al g adual p og ess” (Bolsen 2025; Anonymous
2025).
Violence s Non-Violence
I is now commonplace o hose pushing o change o ou
hei adhe ence o non- iolence as a ma k o alidi y and su-
pe io mo al s anding (Dahlum, Pinckney, and Wig 2023).
Fo example, when desc ibing Google wo ke s’ walkou o
d aw a en ion o Google’s mishandling o sexual ha ass-
men claims and unequal pay, E e (2018) highligh s he
ci ili y and peace ul na u e o he p o es as a posi i e:
I helps ha hese aces a e no ang y, bu mos ly
calm, and ha he walkou s a e la gely peace ul and
ci ilised. Things look e y di e en o ecen p o es s
by Ube d i e s, o example, who ha e s aged mo e
ang y demons a ions, wa ing colou ul signs and
shou ing slogans.
Likewise, adhe ence o non- iolence o en ex ends o p op-
e y des uc ion. As highligh ed by Malm (2021), pas ac-
ions ha e gone so a as o equi ing e e y pa icipan o
abide by a solemn pledge o no damage machines o in as-
uc u e. Adhe ence o non- iolen app oaches also has aca-
demic p oponen s (S ephan and Chenowe h 2008; Neps ad
2011) hough much o his wo k has been a acked by la e
wo k o inaccu acies in analysis and che y-picking o da a
(Gelde loos 2020). This lioniza ion o non- iolence (and
dis ega d o iolen app oaches) has also p opoga ed wi hin
he ech ac i is communi y gene ally, and he AI e hics
communi y speci ically (e.g., h ough ci a ions o S ephan
and Chenowe h (2008) by Boag e al. (2022) as hey ocus
solely on non- iolen collec i e ac ions). While he iew ha
non- iolen ac ions a e mo e e ec i e han iolen ac ions a
achie ing goals is mains eam8, his iew is no uni o mly
sha ed:
I is s ange and s iking ha clima e change ac-
i is s ha e no commi ed any ac s o e o ism. A e
all, e o ism is o he indi idual by a he mode n
8See sec ion “The Need Fo a Radical Flank” which discusses
how, despi e hei signi ican con ibu ions o he Ci il Righ s
Mo emen , Ma in Lu he King J . is o en lionized while el-Hajj
Malik el-Shabazz (i.e., Malcolm X) is la gely o e looked.
wo ld’s mos e ec i e o m o poli ical ac ion, and cli-
ma e change is an issue abou which people eel jus
as s ongly as abou , say, animal igh s. – Lanches e
(2007)
In ac , many ac i is s and academics a gue ha iolen
ac s (e.g., p ope y des uc ion) can be, when engaged wi h
app op ia ely, success ul in di ec ly achie ing hei goals o
u he enabling he success o non- iolen ac o s (Malm
2021) (see: “The Need Fo a Radical Flank” below). His-
o ically, we ha e seen he su age es, he MK in Sou h
A ica, black ac i is s in he U.S. and many o he g oups
success ully use iolence o wo k owa ds achie ing hei
goals. Mo e ecen ly, we’ e seen he success o UK’s Pales-
ine Ac ion in closing down mo e a ms ac o ies h ough di-
ec iolen ac ion (sabo age) han o e a yea o popula
mass p o es s (Al Mayadeen English 2024b,a).
As a gued by Malm (2021), mos people a e likely o
ag ee ha “ iolence is p ima acaie bad”, bu o dis ega d
i as a ma e o cou se in all ci cums ances is bad s a e-
gic hinking. A mo e ealis ic iew on he use o iolence
is ha i is a highly impac ul ool ha should only be used
a e in ense s a egic hinking ha assesses he cul u e and
con ex s. Fo example, i was only a e he pe cei ed ailu e
o non- iolence ha Nelson Mandela s a ed ha “we will
ha e o econside ou ac ics. In my mind we a e closing a
chap e on his ques ion o a non iolen policy” (F eedman
2015). Fo a mo e ecen example on he public suppo o
iolen ac ion, conside he o e whelmingly bi-pa isan sup-
po and sympa hy o Luigi Mangione ollowing his alleged
assassina ion o a heal hca e insu ance company’s CEO (Su-
ciu 2024).
Public s P i a e Ac ions
Wo ke s aking ac ions need o choose be ween ad e ising
hei ac ion, doing i publicly wi hou ad e ising, and doing
i disc ee ly. Public ac ions inc ease sc u iny and possible
pe sonal epe cussions, as well as he impac . Public ac ions
can also se e o inspi e o he s o ac , he impac o which is
likely much g ea e han he impac p i a e ac s ha a single
pe son could execu e. Fu he mo e, he mo e people aking
pa in an ac ion, he g ea e he impac and ha de i will be
o igno e o c ush dissen . Fo example, i ’s i ial o i e and
eplace wo p o es ing wo ke s, bu much ha de o eplace
2000. Likewise, ail-sa es may p e en sabo age caused by
an indi idual bu would likely cease o unc ion i he e was
wide pa icipa ion om many wo ke s.
A he same ime, he e can be many bene i s o aking
ac ions disc ee ly. I one is in a p eca ious posi ion (e.g., on
a isa o suppo ing mul iple dependen s), p i a e ac ions
allow hem o posi i ely impac hei company while mini-
mizing isk. The e a e also ce ain ac ions ha would only
wo k i execu ed disc ee ly. Fo example, ac ions which seek
o su ep i iously d ain he esou ces o companies (e.g., hi -
ing lowe quali y candida es, hi ing poli ically aligned s a ,
and pushing o he s o qui ) would no be e ec i e i exposed
publicly.
Posi i e s Nega i e Ac s
Wo ke s a companies end o ha e di e en poli ics han
he owne ship class (Fe ens ein 2015). As such, hei abil-
i y o in luence he decisions made by managemen is o en
limi ed. Thus, wo ke s o en ha e o ely on dis up ions and
hal ing he no mal ope a ion o he company (i.e., nega i e
ac ions). While nega i e ac ions ha e been e y success ul in
he pas , in a sys em d i en o main ain he s a us-quo, such
ac ions a e no gua an eed o esul in posi i e changes. Tha
is, i all ech co po a ions pa ake in he same immo al ac-
i i ies, e en i wo ke s in all companies engage in ac ions,
he likelihood o companies changing hei ac i i ies is low.
Thus, o d ama ically inc ease he cos o main aining he
s a us-quo, he e should be companies ha exis as mo al al-
e na i es o hose being a ge ed by nega i e ac ions. Build-
ing hese ins i u ions is a e y impo an posi i e ac ion
which we do no belie e ge s enough a en ion among ech
wo ke s. Ins ead o jus qui ing a Big Tech company o join
ano he Big Tech company — which can s ill be a cou a-
geous mo al s ance o ake — qui ing a company o c ea e
a compe i o wi h he in en ion o punishing hem o hei
immo al ac ion c ea es a eal cos o hei ac ions. Now in-
s ead o appealing o he mo als o hose who un companies
and hide behind he excuse o sha e-holde alue o main ain
he s a us-quo, nega i e ac ions will incu a quan i iable cos
ha can be a oided.
Changing om wi hin
A common coun e -a gumen used by hose eluc an o
qui is ha , by being pa o he sys em, wo ke s will ha e
mo e access o in o ma ion, people, and knowledge ha will
make hem mo e e ec i e when pushing o change. Tha is,
change can s ill be achie ed om wi hin a company.
Fo example, wo ke s on he inside can bene i he mo e-
men h ough bai ing and policing bad ac o s. Bai ing col-
leagues who align wi h he ha m ul p ac ices ha he com-
pany p ac ices and espouses in o sha ing hings ha di ec ly
iola e company policy o local laws can esul in hese col-
leagues being epo ed, penalized, and o ced o sel -censo .
The bene i s o such in e nal p essu e a e wo- old: a) penal-
ize bad ai h ac o s while building communi y among hose
campaigning owa ds mo e mo al company alues, and b)
e oding us and ha mony wi hin he company cul u e. In-
e nal ac ions can also cause cowo ke s o also eel mo ally
obliged o ac . Likewise, wo ke pe i ions we e his o ically
mo e impac ul han pe i ions om ou side s.
On he o he hand, many wo ke s inside companies eel
ha , as small cogs inside a massi e machine, hei indi idual
ac ions ha e li le o no e ec . Meaning ul change is di icul
as companies would no keep keep employees who pose any
eal h ea o hei business model o hei ac ions.
Bo h iews ha e u h in hem. The loss o access ( o in-
o ma ion, people, and decision make s) can educe he e -
ec i eness o one’s ac ions. Likewise, as companies no ice
he inc easing poli iciza ion o hei wo k- o ce, hey con-
inue o change in e nal policies, s uc u es, and con ac s
(Pe igo 2024) o e ec i ely neu e any in e nal wo ke dis-
sen . In such ins ances, he bene i s o access a e g ea di-
minished, and mo e d as ic ac ions may be needed o e ec
change.
The Need Fo a Radical Flank
Con empo a y wo ke s ha e demons ably lea ned om he
expe iences o hei p edecesso s. To sus ain momen um,
connec ions, and knowledge gained du ing pe iods o col-
lec i e ac ion (e.g., walkou s o p o es s), which his o ically
ended o dissipa e a e an ini ial su ge, wo ke s ha e now
es ablished endu ing o ganiza ions ha acili a e knowledge
ans e and con inui y (Ma x 2024). Consequen ly, he dis-
missal o indi idual ac i e wo ke s is less likely o hal
mo emen s en i ely. Fu he mo e, wo ke s ha e cul i a ed
in e -company ne wo ks o os e solida i y and dissemina e
lessons lea ned (Ma x 2024).
Howe e , ech wo ke ac i ism cu en ly lacks a adical
lank. While he e exis adicals in ech (e.g., neo-luddi es
(D ies, Luyckx, and Bogae 2024)), hey a e no a lank o
cu en ac i is s. Haines (1984) highligh s how many pas
mo emen s “di ide in o ‘mode a e’ and ‘ adical’ ac ions
du ing hei de elopmen , ha ing been obse ed in he U.S.
labou mo emen , women’s mo emen , an i-nuclea mo e-
men , and he black e ol in he Uni ed S a es”. In he li -
e a u e, he adical lank adop s a mo e maximalis goal and
mo e adical ac ics (Tompkins 2015), hus p esen ing he
mode a e lank as a easonable al e na i e (Malm 2021).
Pe social mo emen heo y, he lack o ac ionalism
wi hin exis ing ech ac i is o ganiza ions (i.e., no adical
lank), sugges s his social mo emen as being in i s in ancy.
This unde s anding can help wo ke s si ua e hemsel es wi h
he con ex o pas s uggles and enable hem o assess which
s a egies a e mo e app op ia e.
Second, unde s anding he inhe en his o ical ension be-
ween mode a e and adical lanks will enable hem o igh
uni ed o a cause. The ole o he mode a es is no become
adical. Ins ead, he mode a e lank should p o-ac i ely wo k
o posi ion i sel o bene i om he spikes in in e es esul -
ing om adical ac ions (Fo ie 2025) while making su e
no o p eemp i ely unde cu any o he ac ions o a adical
lank by undue dispa agemen adical ac ics. As desc ibed
by Malm (2021) o he Black Ci il Righ s s uggle, a jailed
Ma in Lu he King “could signal ha i channels o non-
iolence emained closed, [...] neg eos unques ionably will
look o un ied and pe haps less esponsible leade s – no-
ably Malcolm X” (who ep esen ed he adical lank). He
concludes ha he ci il igh s mo emen ’s mode a e lank
was able o win he Ac o 1964 because, ela i e o he ad-
ical lank, i appea ed o be he lesse o wo e ils.
The li e a u e on he e ec i eness o adical lanks is
mixed. Tompkins (2015), s udying he e ec s o iolen ad-
ical lanks, inds ha hey esul ed in inc eased ep ession
bu we e no “necessa ily de imen al” o p og ess. Simp-
son, Wille , and Feinbe g (2022), like Haines (1984), ound
ha adical lanks inc ease suppo and pe cep ion o mode -
a e lanks. An issue wi h much o he academic li e a u e on
adical lanks is hei conse a i e na u e, aming he adi-
cal lank as doomed o ailu e and only exis ing o u he
he mode a es. Fu he mo e, much o he li e a u e ocuses
Ac ion Risk Impac Impac Immedia e
Violence Towa ds In as uc u e 0.975 0.825 0.925
Unionize 0.4 0.75 0.025
S iking 0.3 0.625 0.75
Dis up i e Wo kplace P o es s 0.625 0.575 0.75
Leaking Publicly 0.575 0.55 0.45
Sabo age in e nal sys ems 0.6 0.55 0.75
Build Compe ing P oduc s -0.3 0.525 -0.075
Sabo age p oduc s 0.675 0.5 0.575
Violence Towa ds O he s 0.9 0.375 0.4
Lobbying -0.35 0.325 -0.125
Qui (ex e nal p o es ) 0.425 0.3 0.35
Leaking Anonymously -0.359 0.25 0.4
Boyco (nega i e) -0.425 0.225 0
Re use o build 0.225 0.2 0.4
Public P o essional P o es s 0.375 0.175 0.375
Lawsui s 0.4 0.15 -0.25
Boyco (posi i e) -0.475 0.05 -0.125
Hi ing Poli ically Aligned S a -0.6 -0.025 -0.2
Qui (in e nal p o es ) 0.15 -0.075 0.1
In e nal P o es s (ac i e) 0.175 -0.1 0
Push o qui ing -0.2 -0.15 -0.075
(I i ed) Sue 0.25 -0.2 -0.375
Educa ional campaigns/ each-ins -0.375 -0.2 -0.2
In e nal P o es s (passi e) 0.1 -0.2 -0.075
C ea e assis i e ools/ esou ces -0.65 -0.25 -0.175
Engage wi h ac i is /ci il igh s g oups -0.525 -0.25 -0.075
Repo ing o Regula o y Body -0.375 -0.3 -0.375
Hi ing lowe quali y candida es -0.525 -0.35 -0.5
Elonga e bu eauc a ic p ocesses -0.375 -0.425 -0.35
Violence Towa ds Sel 0.854 -0.425 -0.1
Publicize Ac ions -0.3 -0.475 -0.3
Repe a i e Ac ions -0.75 -0.5 -0.5
Minimal p oduc i i y -0.475 -0.525 -0.475
Bai ing Reac ions -0.25 -0.575 -0.35
Engage wi h HR -0.025 -0.575 -0.475
Public P o es s -0.65 -0.6 -0.35
Qui (quie ly) -0.05 -0.75 -0.725
Table 3: Risk, Impac , and Immedia e Impac as a ed by ou anno a o s.