Concerns with Intentional Insights

A re­cent face­book post by Jeff Kauf­man raised con­cerns about the be­hav­ior of In­ten­tional In­sights (InIn), an EA-al­igned or­ga­ni­za­tion headed by Gleb Tsipursky. In dis­cus­sion aris­ing from this, a num­ber of fur­ther con­cerns were raised.

This post sum­ma­rizes the con­cerns found with InIn. It also notes some con­cerns which were mis­taken and un­founded, and facts that arose which re­flect well on InIn.

This post was con­tributed to by Jeff Kauf­man, Gre­gory Lewis, Oliver Habryka, Carl Shul­man, and Claire Za­bel. They dis­close rele­vant con­flicts of in­ter­est be­low.


1 Ex­ag­ger­ated claims of af­fili­a­tion or en­dorse­ment
1.1 Kerry Vaughan of CEA
1.2 Giv­ing What We Can (GWWC)
1.3 An­i­mal Char­ity Eval­u­a­tors (ACE)
2 Astro­tur­fing
2.1 The In­ten­tional In­sights blog
2.2 The Effec­tive Altru­ism fo­rum
2.3 LessWrong
2.4 Face­book
2.4.1 Solic­it­ing up­votes and deny­ing it
2.4.2 Not dis­clos­ing paid sup­port
2.5 Ama­zon
3 Mislead­ing figures
4 Du­bi­ous prac­tices
4.1.1 Paid con­trac­tors’ ex­pected ‘vol­un­teer­ing’
4.1.2 Fur­ther de­tails re­gard­ing con­trac­tor ‘vol­un­teer­ing’
4.2 Ama­zon best­sel­ler
5 In­flated so­cial me­dia im­pact
5.1 Face­book
5.2 The Life You Can Save dona­tions
5.3 Twit­ter
5.4 Pin­ter­est
5.5 Pre­sen­ta­tions of me­dia ar­ti­cle traf­fic and reach
5.5.1 TIME ar­ti­cle
5.5.2 Huffing­ton Post
6 Mis­taken/​Un­fair ac­cu­sa­tions
6.1 Sup­posed lin­ear­ity of Twit­ter fol­lower in­crease
6.2 Ob­jec­tions to In­ten­sional In­sights staff ‘lik­ing’ In­ten­tional In­sights con­tent
6.3 ‘Paid likes’ from click­farms
7 Pos­i­tives
7.1 Jon Be­har
7.2 Ad­di­tional dona­tions
7.3 Place­ment of ar­ti­cles in TIME and the Huffing­ton Post
8 Policy re­sponses from InIn
8.1 Post-crit­i­cism con­flict-of-in­ter­est policy
8.2 Post-crit­i­cism Face­book boost­ing
9 Dis­clo­sures
10 Re­sponse com­ments from Gleb Tsipursky

1. Ex­ag­ger­ated claims of af­fili­a­tion or endorsement

In­ten­tional In­sights claims ‘ac­tive col­lab­o­ra­tions’ with a num­ber of Effec­tive Altru­ist groups in its The­ory of Change doc­u­ment which was on its “About” page (Au­gust 21, 2016).

In a num­ber of cases InIn makes use of the name of an effec­tive al­tru­ist or­ga­ni­za­tion with­out ask­ing for that or­ga­ni­za­tion’s con­sent, based on minor in­ter­ac­tions such as the or­ga­ni­za­tion an­swer­ing ques­tions about web traf­fic. From the ‘Effec­tive Altru­ism im­pact of In­ten­tional In­sights’ doc­u­ment (Au­gust 19, 2016):

As de­tailed be­low, we ob­serve that af­ter learn­ing of such claims and use of their names, some of these groups had asked InIn to stop. Yet even in some of these cases InIn had not al­tered the men­tions in its pro­mo­tional ma­te­ri­als months later. Tsipursky also does not ap­pear to have adopted a prac­tice of check­ing with or­ga­ni­za­tions be­fore us­ing their names in InIn pro­mo­tional ma­te­ri­als.

1.1. Kerry Vaughan of CEA

Tsipursky pre­vi­ously posted notes from a Skype con­ver­sa­tion with Kerry Vaughan with­out his con­sent, and sug­gested he had en­dorsed In­ten­tional In­sights where he had not:

Tsipursky later apol­o­gized, ed­ited the post, and said he had up­dated. Yet he later en­gaged in similar be­hav­ior (see sec­tions 1.2 and 1.3 be­low).

1.2. Giv­ing What We Can (GWWC)

Gleb has taken the Giv­ing What We Can pledge, and con­tributed an ar­ti­cle on the Giv­ing What We Can blog on De­cem­ber 23, 2015. He also men­tioned and linked to GWWC in his ar­ti­cles el­se­where.

Michelle Hutch­in­son, Ex­ec­u­tive Direc­tor of Giv­ing What We Can, wrote to Tsipursky in May 2016 ask­ing him to cease “claiming to be sup­ported by Giv­ing What We Can.” How­ever, the use of Giv­ing What We Can’s name as an ‘ac­tive col­lab­o­ra­tion’ was not re­moved from In­ten­tional In­sights’ web­site, and re­mained in both of the above InIn doc­u­ments as of Oc­to­ber 15, 2016.

1.3. An­i­mal Char­ity Eval­u­a­tors (ACE)

In the InIn im­pact doc­u­ment Tsipursky quotes Leah Edger­ton of ACE:

Erika Alonso of ACE sub­se­quently made the fol­low­ing state­ment:

2. Astroturfing

Astro­tur­fing is giv­ing the mis­lead­ing im­pres­sion of un­af­fili­ated (“grass­roots”) sup­port. In GiveWell’s first year its cofounders en­gaged in as­tro­tur­fing, and this was taken very se­ri­ously by its board. Among other re­sponses, the GiveWell board de­moted one of the co-founders and fined both $5,000 each. Tsipursky ex­pressly claimed not to en­gage in as­tro­tur­fing:

How­ever, as­tro­tur­fing is wide­spread across the In­ten­tional In­sights so­cial me­dia pres­ence (doc­u­mented in the sec­tions be­low). Tsipursky did qual­ify his state­ment with “we are not ask­ing peo­ple to do these sorts of ac­tivi­ties in their paid time”, but lack of pay­ment isn’t enough to pre­vent mis­lead­ing peo­ple about the na­ture of the sup­port. In any case, the dis­tinc­tion be­tween con­trac­tors’ paid and un­paid time is blurry (see sec­tion 4.1.1).

2.1. The In­ten­tional In­sights blog

Paid con­trac­tors for In­ten­tional In­sights leave com­pli­men­tary re­marks on the In­ten­tional In­sights blog, and the In­ten­tional In­sights ac­count replies with grat­i­tude, as if the com­ments were by strangers. At no stage do they dis­close the fi­nan­cial re­la­tion­ship that ex­ists be­tween them. In the screen­shot be­low (source, Candice, John, Beatrice, Jojo, and Shyam are all In­ten­tional In­sights con­trac­tors.

The most re­cent ex­am­ples of this hap­pened in late Au­gust 2016, af­ter the ini­tial post and dis­cus­sion with Tsipursky on Jeff’s Face­book wall, and dur­ing the draft­ing of this doc­u­ment.

2.2. The Effec­tive Altru­ism forum

Tsipursky has done the same thing on the Effec­tive Altru­ism fo­rum. Here is one in­stance (note that “Nyor” also goes by “Jojo”):

Here is an­other ex­am­ple (note that “An­tho­nye­muobo” is a pro­fes­sional han­dle used by one of Tsipursky’s ac­knowl­edged con­trac­tors, “Sar­gin”):

2.3. LessWrong

Tsipursky posted a link to some of his wife and InIn co-founder’s writ­ing in Fe­bru­ary 2016, with­out not­ing this con­nec­tion:

This is a minor lapse, one which Gleb claimed to have learned from and up­dated. Yet similar be­hav­ior con­tinued:

In March 2016, In­ten­tional In­sights’ con­trac­tors cre­ated ac­counts and started post­ing non-spe­cific praise on Tsipursky’s LessWrong posts:

Th­ese are all peo­ple Tsipursky pays, but none of them ac­knowl­edged it in their com­ments or their posts in the wel­come thread. Ad­di­tion­ally, Tsipursky did not ac­knowl­edge this re­la­tion­ship when he thanked them for their re­marks.

LessWrong user gjm pointed out that this was mis­lead­ing, and Tsipursky ac­knowl­edged this was a prob­lem and com­mented on Sar­gin’s wel­come post:

While Tsipursky knew both Beatrice Sar­gin and Alex Wences­lao had posted similar com­ments, since he had replied to them, he waited for these to be dis­cov­ered and pointed out be­fore act­ing:

This hap­pened a third time, with JohnC2015:

2.4. Facebook

2.4.1. Solic­it­ing up­votes and deny­ing it

Tsipursky claimed “when I make a post on the EA Fo­rum and LW I will let peo­ple who are in­volved with InIn know about it, for their con­sid­er­a­tion, and ex­plic­itly don’t ask them to up­vote”:

In the com­ment Tsipursky de­nies so­lic­it­ing up­votes, and de­mands that ac­cu­sa­tions that he did be sub­stan­ti­ated or with­drawn. Six hours later some­one re­sponded with a screen­shot of a post Tsipursky had made to the In­ten­tional In­sights In­sid­ers group show­ing Tsipursky so­lic­it­ing up­votes:

Tsipursky’s re­sponse, a cou­ple hours later in the same thread:

Tsipursky ei­ther gen­uinely be­lieved posts like the above do not ask for up­votes, or he be­lieved state­ments that are mis­lead­ing on com­mon-sense in­ter­pre­ta­tion are ac­cept­able pro­vid­ing they are ar­guably ‘true’ on some ten­den­tious read­ing. Nei­ther is re­as­sur­ing. [He sub­se­quently con­ceded this was ‘less than fully forth­com­ing’.]

2.4.2. Not dis­clos­ing paid support

In­ten­tional In­sights pro­posed pro­duc­ing EA T-Shirts, and re­ceived mul­ti­ple crit­i­cisms. Tsipursky claimed he had run the de­sign by mul­ti­ple peo­ple. Again, Tsipursky did not dis­close that at least five of them were peo­ple he pays:

2.5. Amazon

Tsipursky’s con­trac­tor posted a 5-star re­view for his self-help book on Ama­zon with­out dis­clos­ing the af­fili­a­tion:

Tsipursky emailed copies of his self-help book to In­ten­tional In­sights vol­un­teers, in­clud­ing con­trac­tors, who re­sponded by post­ing 5-star re­views on Ama­zon:

He later fol­lowed up with:

This is true but in­com­plete: the 8th re­view is by As­raful Is­lam, a vol­un­teer af­fili­ated to In­ten­tional In­sights.

Another In­ten­tional In­sights af­fili­ate, un­paid at that time but now a paid vir­tual as­sis­tant, Elle Ac­quino, posted an­other 5-star re­view, not in the top 10. In that re­view, how­ever, the con­nec­tion to Tsipursky and his non­profit in­sti­tute was dis­closed.

3. Mislead­ing figures

In De­cem­ber 2015 and Jan­uary 2016, Tsipursky re­peat­edly claimed that his ar­ti­cles were shared thou­sands of times as ev­i­dence of the effec­tive­ness of his ap­proach. In fact, he had been re­port­ing Face­book ‘likes’ and all views on Stum­ble­upon as shares, greatly ex­ag­ger­at­ing the ex­tent of so­cial me­dia en­gage­ment.

The ini­tial point re­flected a com­mon is­sue with the in­ter­pre­ta­tion of so­cial me­dia ac­tivity coun­ters on web­sites. After this was ex­plained to him Tsipursky claimed to have up­dated on the cor­rec­tion. How­ever, a June 2016 doc­u­ment on In­ten­tional In­sight’s Effec­tive Altru­ism Im­pact again re­ported views as shares, ex­ag­ger­at­ing shar­ing by many times.

4. Du­bi­ous practices

4.1.1. Paid con­trac­tors’ ex­pected ‘vol­un­teer­ing’

Tsipursky only takes on con­trac­tors who spend at least two hours “vol­un­teer­ing” for In­ten­tional In­sights for each paid hour:

In a fol­low-up dis­cus­sion, Tsipursky sug­gested that con­trac­tors could tem­porar­ily re­duce their vol­un­teer hours in spe­cial cir­cum­stances, but he would not af­firm that con­trac­tors would be al­lowed to sim­ply say no to “vol­un­teer­ing”:

Depend­ing on the na­ture of the vol­un­teer work, this re­quire­ment seems po­ten­tially un­eth­i­cal, effec­tively re­quiring that con­trac­tors do three times as much work for a fixed amount of money. We also sug­gest this re­la­tion­ship un­der­mines the dis­tinc­tion Tsipursky offers be­tween ‘paid’ and ‘vol­un­teer time’ and the defence that the pro­mo­tion his con­trac­tors un­der­take on his be­half is in­nocu­ous as it oc­curs in their ‘vol­un­teer time’.

4.1.2. Fur­ther de­tails re­gard­ing con­trac­tor ‘vol­un­teer­ing’

Sub­se­quent to the prepa­ra­tion of the above sec­tion Tsipursky pro­vided ad­di­tional in­for­ma­tion about how he came into con­tact with con­trac­tors, their dona­tions, prior un­paid vol­un­teer­ing, wages, and other in­for­ma­tion as ev­i­dence of gen­uine sup­port. They provide that, but also sup­port con­cerns re­gard­ing link­age of paid and un­paid work and fi­nan­cial in­ter­ests.

Tsipursky states the fol­low­ing re­gard­ing ini­tial meet­ings and hiring:

Tsipursky stated the fol­low­ing re­gard­ing the length of un­paid vol­un­teer­ing prior to the first paid work:

He also notes dona­tions by con­trac­tors, im­ple­mented by re­duc­ing their paid hours or paid hour wage rate, as ev­i­dence of gen­uine sup­port:

I have pointed out many times that there is plenty of ev­i­dence show­ing that those folks who do con­tract­ing are pas­sion­ate en­thu­si­asts for InIn. Let’s take the ex­am­ple of John Chavez, who the doc­u­ment brought up. He chose to re­spond to a fundrais­ing email to our sup­porter list­serve in June 2016 – long be­fore Jeff Kauf­man’s origi­nal post – by donat­ing $50 per month to InIn out of his $300 monthly salary:

This is big­ger than a typ­i­cal GWWC mem­ber, at over 15% of his an­nual in­come. Let me re­peat – he vol­un­tar­ily, out of his own vo­li­tion in re­sponse to a fundrais­ing that went out to all of our sup­port­ers, chose to make this dona­tion. Just to be clear, we send out fundrais­ing let­ters reg­u­larly, so it’s not like this was some spe­cial oc­ca­sion. It was just that – as he said in the let­ter – it hap­pened to fall on the 1-year an­niver­sary of him join­ing InIn and he felt in­spired and moved by the mis­sion and work of the or­ga­ni­za­tion to give.

Be­fore you go say­ing John is unique, here is an­other screen­shot of a dona­tion from an­other con­trac­tor who in Oc­to­ber 2015, in re­sponse to a fundrais­ing email, made a $10/​month dona­tion:

Again, vol­un­tar­ily, out of her own vo­li­tion, she chose to make this dona­tion.

Tsipursky also in­di­cates that paid and un­paid hours by con­trac­tors con­sti­tute only a minor­ity of work hours at InIn, with most hours con­tributed by vol­un­teers with­out fi­nan­cial com­pen­sa­tion:

Re­gard­ing wages and re­quire­ment/​ex­pec­ta­tions of un­paid vol­un­teer­ing, Tsipursky wrote the fol­low­ing:

The Up­work (formerly known as Odesk) free­lancer mar­ket­place on which con­trac­tors are hired has a min­i­mum wage of $3.00 per hour. Com­bined with the ex­pected un­paid vol­un­teer­ing the typ­i­cal wage would be $1.00, 1/​3rd of the min­i­mum for the plat­form.

John is given as an ex­am­ple of a higher paid con­trac­tor at $7.5 per hour. How­ever, this is com­bined with 3 hours of un­paid vol­un­teer­ing for each paid hour, rather than 2, for a com­bined wage of $1.875 per hour, prior to his dona­tion of 13 of that wage.

In effect, the ex­pec­ta­tion of vol­un­teer­ing sys­tem­at­i­cally cir­cum­vents the Up­work min­i­mum wage for con­trac­tors. How­ever, it should be noted that the Up­work min­i­mum wage is a cor­po­rate policy, and not a na­tional or lo­cal la­bor law. Con­trac­tors in low-in­come coun­tries may be earn­ing sub­stan­tially more than the lo­cal min­i­mum wages or av­er­age in­comes. For ex­am­ple, ac­cord­ing to Wikipe­dia the hourly min­i­mum wage in US dol­lars at nom­i­nal ex­change rates is $0.54 in Nige­ria. In the Philip­pines min­i­mum wages vary by lo­ca­tion and sec­tor, but Wikipe­dia lists a range of ~$0.6-$1.2 per hour for non-agri­cul­tural work­ers, with the lat­ter group in the cap­i­tal of Manila. So the wage per com­bined (paid+vol­un­teer) hour of work would not ap­pear to be in con­flict with le­gal min­i­mum wages in con­trac­tors’ ju­ris­dic­tions. Fur­ther­more in a num­ber of these ju­ris­dic­tions the min­i­mum wage is closer to the me­dian wage, and un­em­ploy­ment is high.

Re­gard­ing the link be­tween paid and un­paid hours, Tsipursky de­scribes it as an in­for­mal un­der­stand­ing:

In ag­gre­gate the ad­di­tional state­ments provide ev­i­dence of pre-ex­ist­ing sup­port for InIn from new con­trac­tors. How­ever, they also con­firm a link­age of paid and un­paid la­bor, and con­trac­tor fi­nan­cial in­ter­ests in pro­mo­tional ac­tivity oc­cur­ring dur­ing ‘vol­un­teer’ hours.

4.2. “Best-sel­l­ing au­thor”

Tsipursky in­cludes be­ing a ‘best-sel­l­ing au­thor’ in his stan­dard bio. For ex­am­ple, on his Pa­treon:


And on his Ama­zon au­thor page:

Nor­mally, a reader would take “best-sel­l­ing au­thor” to mean hit­ting a ma­jor best-sel­ler list like the New York Times, which in­di­cates that very many peo­ple have de­cided to buy the book, and is a hard sig­nal to fake. In Tsipursky’s case, “best-sel­l­ing au­thor” means that his book was very briefly the top sel­ler in a sub-sub-cat­e­gory of Ama­zon. Fur­ther, he re­ports offer­ing his book for free and en­courag­ing friends and con­trac­tors to down­load and re­view it. In its first two weeks the book sold 50 copies at $3 each. Cu­mu­la­tively it has sold 500 copies at $3 each, and been down­loaded 3500+ times free. In con­trast, NYT best­sel­ler sta­tus re­quires thou­sands of sales over the first week. Ama­zon best­sel­ler sta­tus is calcu­lated hourly by cat­e­gory: in small cat­e­gories three pur­chases in an hour can win the #1 best­sel­ling au­thor la­bel.

Many of those giv­ing the book 5 star re­views are so­cial con­tacts of Tsipursky, some of them paid or vol­un­teer In­ten­tional In­sights staff, but do not dis­close this as­so­ci­a­tion (see sec­tion 2.5).

As of Au­gust 22, 2016 the book is ranked as fol­lows:

In light of this, call­ing one­self a ‘best­sel­ling au­thor’ on this sort of perfor­mance is po­ten­tially mis­lead­ing.

We note that the prac­tice of claiming best­sel­ling au­thor sta­tus us­ing best­sel­ler lists that in­volve very small ac­tual sales may be wide­spread. This does not, how­ever, pre­vent it from be­ing mis­lead­ing or con­tro­ver­sial. For ex­am­ple, when Brent Un­der­wood at­tained Ama­zon best-sel­ler sta­tus us­ing a few dol­lars in less than an hour with a book that was sim­ply a pic­ture of his foot, me­dia cov­er­age gen­er­ally sug­gested that this high­lighted a prob­le­matic prac­tice.

5. In­flated so­cial me­dia impact

5.1. Facebook

Tsipursky has cited so­cial me­dia en­gage­ment as ev­i­dence of im­pact. How­ever, in many cases it ap­pears that this en­gage­ment is illu­sory. In the case of Face­book, it ap­pears to have re­sulted from paid Face­book post boost­ing, which led to hun­dreds of likes on posts from click­farms, in a pro­cess de­scribed by Ver­i­ta­sium: click­farm ac­counts like enor­mous num­bers of things they have not been di­rectly paid to like in or­der to ma­nipu­late Face­book’s al­gorithms. Face­book boost­ing sys­tem­at­i­cally at­tracts these click­farm ac­counts, a risk which is ex­ac­er­bated by boost­ing to re­gions where click­farms are lo­cated (al­though click­farms also have fake ac­counts pur­port­ing to be from all around the world).

In the case of InIn posts, InIn paid for that boost­ing. In Fe­bru­ary 2016, Tsipursky ar­gued that this was re­sult­ing in gen­uine en­gage­ment and reach:

For a num­ber of InIn blog posts with large num­bers of likes (for ex­am­ple 318 for this re­cent one) these likes ap­pear to be pri­mar­ily the re­sult of click­farms. Ac­counts lik­ing this post like vast num­bers of dis­parate things. Here are some ran­dom se­lec­tions from the mid­dle of the list of that post:

There is fur­ther cir­cum­stan­tial ev­i­dence: the likes are of­ten from ac­counts in low-in­come coun­tries with sub­stan­tial click­farm op­er­a­tions. Tsipursky defended this as co­in­ci­den­tal over­lap caused by In­ten­tional In­sight’s tar­get­ing of low-in­come coun­tries, how­ever coun­tries with similar de­mo­graph­ics with­out large click farm op­er­a­tions are not well rep­re­sented.

In ar­gu­ing for the im­pact of his writ­ing, Tsipursky cited a post on the TLYCS blog that got 500 likes in its first day on the TLYCS blog while typ­i­cal posts got 100-200 likes:

How­ever, this ap­pears to also be a case of Face­book ad boost­ing elic­it­ing en­gage­ment from click­farms, this time by a former TLYCS em­ployee (sub­se­quently asked to stop by TLYCS) rather than InIn, ac­cord­ing to this state­ment from TLYCS’ Jon Be­har:

The pro­files con­tribut­ing the likes and whose pro­files show no other en­gage­ment with TLYCS, or with EA ideas:

After Jeff Kauf­man raised con­cerns about the pat­tern of Face­book likes in Fe­bru­ary 2016, Tsipursky doesn’t seem to have looked into the is­sue fur­ther prior to the Au­gust 2016 dis­cus­sion, when out­side ob­servers pro­vided in­dis­putable ev­i­dence and ex­plained the role of boost­ing in gen­er­at­ing click­farm likes. While the boost­ing-click­farm link is coun­ter­in­tu­itive, the lack of any other en­gage­ment by the click­farm­ers was ap­par­ent both be­fore and af­ter the rais­ing of con­cerns in Fe­bru­ary. Failure to ex­am­ine the in­effec­tive­ness of these so­cial me­dia chan­nels, even af­ter con­cerns were raised, raises ques­tions about InIn’s prac­tices as an out­reach and con­tent mar­ket­ing or­ga­ni­za­tion.

5.2. The Life You Can Save donations

In his “Effec­tive Altru­ism im­pact of In­ten­tional In­sights” doc­u­ment (archived copy), Tsipursky claims that con­tent he has pub­lished with The Life You Can Save is able to “reg­u­larly reach an au­di­ence of over 5,000, at least 12% of whom make a dona­tion” sug­gest­ing over 600 dona­tions per ar­ti­cle, based on a refer­ence let­ter from a former TLYCS em­ployee. How­ever, these figures were in­cor­rect, and TLYCS es­ti­mates that the to­tal num­ber of vis­i­tors who landed on Tsipursky’s blog posts at the TLYCS blog was ~3,000 (rather than tens of thou­sands), with dona­tions di­rectly from those page to­tal­ling likely 2-3 (rather than hun­dreds).

While the refer­ence let­ter Tsipurksy cites could eas­ily give that false im­pres­sion, it is im­plau­si­ble in light of other in­for­ma­tion available to him about the im­pact of his pieces. For ex­am­ple, Tsipursky also cites an ar­ti­cle in a ma­jor news out­let as pro­duc­ing two dona­tions to GiveDirectly to­tal­ling $500:

Since two dona­tions is far less than ~600, this “12% of 5,000 views” num­ber was clearly not san­ity checked be­fore be­ing used to ar­gue the case for In­ten­tional In­sights to EAs and in a fundrais­ing doc­u­ment aimed at EAs. It’s pos­si­ble that Tsipursky sim­ply took a sur­pris­ingly good es­ti­mate from a part­ner or­ga­ni­za­tion at face value, but one might ex­pect an ex­pert in mar­ket­ing to in­ves­ti­gate why this chan­nel was perform­ing so much bet­ter than his other chan­nels.

5.3. Twitter

Tsipursky im­plied that his 10k Twit­ter fol­low­ers rep­re­sent or­ganic in­ter­est:

The InIn ac­count is fol­low­ing ap­prox­i­mately as many ac­counts as fol­low it, 11.7k to 11.4k. Oliver ob­served that many of these ac­counts have “100% fol­low-back” in their de­scrip­tions. It seems like they’re offer­ing an ex­change: InIn adds these ac­counts as fol­low­ers, and they add his back in re­turn, or vice versa. This is not an in­di­ca­tion of ac­tual in­ter­est from fans, and these ac­counts have al­most no or­ganic en­gage­ment with InIn such as retweets:

5.4. Pinterest

InIn fol­lows over 20,000 peo­ple on Pin­ter­est, far more peo­ple than fol­low it. As on the InIn Face­book page and Twit­ter, fol­lower en­gage­ment is ex­tremely low, and dom­i­nated by per­sons af­fili­ated with InIn, sug­gest­ing the vast ma­jor­ity of fol­low­ers are not gen­uine.

Ex­am­in­ing the pro­files of fol­low­ers, there ap­pears to be a very high rate of click­farm/​ad­ver­tis­ing ac­counts. Here are 10 ran­domly se­lected InIn Pin­ter­est fol­lower ac­counts. 10 out of 10 ap­pear to be spam/​ad­ver­tis­ing/​click­farm ac­counts:

5.5. Pre­sen­ta­tions of me­dia ar­ti­cle traf­fic and reach

5.5.1. TIME article

In the InIn EA im­pact doc­u­ment we see this:

The doc­u­ment does not make clear that the ar­ti­cle did not ap­pear in the print mag­a­z­ine, so print read­ers would not be ex­posed to it there. On­line, we are left to an­chor on a figure of 65 mil­lion views, with­out any refer­ence to the ac­tual views of the ar­ti­cle (which were tremen­dously lower).

Some­what later in the doc­u­ment we see this:

As an­other ex­am­ple, here are num­bers in a spread­sheet we set up re­cently to track clicks to EA non­profit web­sites from the Time piece we pub­lished.

How­ever, while the ar­ti­cle made the case for GiveWell recom­mended char­i­ties and EA char­ity eval­u­a­tors, only 132 clicks reached those or­ga­ni­za­tions through the ar­ti­cle, 70 of which did not im­me­di­ately bounce, ac­cord­ing to InIn’s traf­fic figures. Speci­fi­cally, in the origi­nal InIn spread­sheet the ‘signed up to newslet­ter or con­verted in other ways″ column had a value of 13 for ACE, and 1 ‘clicked on donate but­ton’

The cor­rected spread­sheet shows a value of 2 rather than 13 for ‘signed up to newslet­ter.’

Thus InIn knew that the product of traf­fic and click-through was very low, sug­gest­ing some com­bi­na­tion of low traf­fic for a piece on Time’s web­site and low click-through rates. How­ever this nega­tive in­for­ma­tion was re­moved from the main text of the doc­u­ment while the 65 mil­lion figure (for all ar­ti­cles on the TIME web­site, in­clud­ing du­bi­ous traf­fic) was made promi­nent.

5.5.2. Huffing­ton Post

The InIn EA im­pact doc­u­ment also in­cluded this dis­cus­sion of a Huffing­ton Post ar­ti­cle:

How­ever, he pro­vided no ev­i­dence of reach­ing new au­di­ences via the place­ment in the HuffIng­ton Post. In­stead, he pro­vided an ex­am­ple of an already sup­port­ive face­book friend, who ap­par­ently en­coun­tered the ar­ti­cle from Tsipursky’s Face­book page, not the Huffing­ton Post.

6. Mis­taken/​Un­fair accusations

6.1. Sup­posed lin­ear­ity of Twit­ter fol­lower increase

It was sug­gested that Tsipursky’s twit­ter page shows sur­pris­ingly lin­ear in­creases in fol­low­ers over time (e.g. +8 fol­low­ers a day for 10 days in a row, which may be in­dica­tive of click-farm­ing. This piece of ev­i­dence is likely mis­taken, as the tool used (share­counter) prob­a­bly lin­early in­ter­po­lates days where they do not record a user’s Twit­ter fol­low­ers, and thus the ap­par­ent lin­ear­ity is an ar­ti­fact.

6.2. Ob­jec­tions to In­ten­sional In­sights staff ‘lik­ing’ In­ten­tional In­sights content

In the course of the origi­nal dis­cus­sion of Jeff’s post on Face­book, nu­mer­ous peo­ple took ex­cep­tion to staff or vol­un­teers ‘lik­ing’ or sup­port­ing InIn con­tent. This crit­i­cism is mis­guided: this is com­mon prac­tice both for non­prof­its gen­er­ally and within the EA com­mu­nity: many EAs af­fili­ated to a given group ‘like’ or share con­tent with­out dis­clos­ing their af­fili­a­tion. Although is­sues around ap­pro­pri­ate dis­clo­sure can be sub­tle, acts like this on so­cial me­dia do not on re­flec­tion seem sig­nifi­cant enough to war­rant dis­clo­sure of in­ter­ests to the au­thors of this doc­u­ment.

6.3. ‘Paid likes’ from clickfarms

In the Fe­bru­ary 2016 dis­cus­sion it was sug­gested the Tsipursky might be di­rectly pay­ing for likes from click­farms. How­ever, as dis­cussed in sec­tion 5.1, while the likes in ques­tion ap­pear to have re­sulted from paid Face­book boost­ing, and to be from click­farms, they were not di­rectly paid for. In­stead, the boost­ing at­tracted click­farm likes through an ac­ci­den­tal pro­cess ex­plained well the linked Ver­i­ta­sium video.

7. Positives

In the course of re­search into and dis­cus­sion around InIn, some facts that re­flect well on InIn were dis­cov­ered. Th­ese are listed be­low. We don’t think this com­prises all ev­i­dence favourable of InIn: the im­pact doc­u­ment, Tsipursky’s post on the EA fo­rum, and the In­ten­tional In­sights web­site offer fur­ther ev­i­dence. (We have not looked at these closely enough to have a view on them.)

7.1. Jon Behar

One TLYCS em­ployee who was worked with Tsipursky on Giv­ing Games says Tsipursky has made helpful in­tro­duc­tions:

Be­har is also quoted in the InIn EA im­pact doc as say­ing:

7.2. Ad­di­tional donations

TLYCS has in­for­ma­tion in­di­cat­ing that Tsipursky’s posts com­bined drove about two or three dona­tions, and the Huffing­ton Post ar­ti­cle re­sulted in to dona­tions to GiveDirectly to­tal­ing $500. Track­ing dona­tions is hard, so this is definitely an un­der­es­ti­mate.

7.3. Place­ment of ar­ti­cles in TIME and the Huffing­ton Post

Tsipursky’s ar­ti­cles in TIME and the Huffing­ton Post got lots of ex­po­sure for EA ideas. Ad­di­tion­ally, be­ing able to get ar­ti­cles placed there is im­pres­sive.

8. Policy re­sponses from InIn

Dur­ing dis­cus­sions with Tsipursky re­gard­ing drafts of this doc­u­ment he men­tioned some InIn policy changes made in re­sponse to the crit­i­cisms. This sec­tion does not re­flect any other changes InIn may have made, pri­mar­ily be­cause we haven’t been able to put in the time to fol­low up on each prac­tice and see whether it has con­tinued. We also note that Tsipursky pro­vided ad­di­tional in­for­ma­tion re­gard­ing Ama­zon sales, con­trac­tor names, and pay­ment prac­tices upon re­quest for this doc­u­ment.

8.1. Post-crit­i­cism con­flict-of-in­ter­est policy

Fol­low­ing the dis­cus­sion un­der Jeff Kauf­man’s post in Au­gust 2016, InIn cre­ated a con­flicts of in­ter­est policy doc­u­ment:

8.2. Post-crit­i­cism Face­book boosting

Tsipursky now states:

Re­gard­ing InIn so­cial me­dia policy, we are mak­ing sure to avoid boost­ing any more posts to click­farm coun­tries. We’re gen­er­ally not boost­ing posts right now to any­one but fans of the page who live in the US and other rich coun­tries. We found we couldn’t ban iden­ti­fi­able click­farm ac­counts from the FB page, un­for­tu­nately, so we’re be­ing re­ally cau­tious about boost­ing posts.

9. Disclosures

Many peo­ple con­tributed to this doc­u­ment, some of them anony­mously. Below are dis­clo­sures from peo­ple who con­tributed sub­stan­tially and want to be clear about any po­ten­tial con­flicts of in­ter­est. None of the in­di­vi­d­u­als be­low con­tributed on be­half of an em­ployer or or­ga­ni­za­tion, and their con­tri­bu­tions should not be taken to im­ply any stance on the part of any or­ga­ni­za­tion with which they are af­fili­ated.

  • Jeff Kauf­man has donated to the Cen­tre for Effec­tive Altru­ism (CEA), 80,000 Hours, and Giv­ing What We Can. He has vol­un­teered for An­i­mal Char­ity Eval­u­a­tors in a very minor ca­pac­ity. His wife, Ju­lia Wise, works for CEA and serves on the board of GiveWell.

  • Gre­gory Lewis has pre­vi­ously worked as a vol­un­teer for Giv­ing What We Can and 80,000 hours. He has donated to Giv­ing What We Can and the Global Pri­ori­ties Pro­ject.

  • Oliver Habryka cur­rently works for CEA, and has been ac­tive in EA com­mu­nity or­ga­niz­ing in a va­ri­ety of roles.

  • Carl Shul­man cur­rently works for the Fu­ture of Hu­man­ity In­sti­tute, and con­sults for the Open Philan­thropy Pro­ject. He pre­vi­ously worked for the Ma­chine In­tel­li­gence Re­search In­sti­tute (MIRI). He has pre­vi­ously done some con­sult­ing and vol­un­teer­ing for the Cen­ter for Effec­tive Altru­ism, es­pe­cially 80,000 Hours. His wife is ex­ec­u­tive di­rec­tor of the Cen­ter for Ap­plied Ra­tion­al­ity and a board mem­ber of MIRI.

  • Claire Za­bel works at the Open Philan­thropy Pro­ject, and serves on the board of An­i­mal Char­ity Eval­u­a­tors. She has donated to a va­ri­ety of EA or­ga­ni­za­tions and has close ties with other peo­ple in the EA com­mu­nity.

10. Re­sponse com­ments from Gleb Tsipursky

Tsipursky has re­sponded in the com­ments be­low: part one, part two, part three.