What Tech Workers Talk about When They Talk about Collective Action
by Anonymous1 and Bao Kham Chau2
Collective action is a key leverage point for technology professionals trying to make change in their industry. With this in mind, we investigate how technology professionals talk about various kinds of collective action (e.g., open letters and unionization efforts). Linking past instances of collective action to discussion threads on the popular social news website Hacker News (HN), we find that HN users’ attention is not distributed equally. Actions carried out by white-collar tech workers at U.S. companies are the most actively discussed. Actions occurring outside the U.S. and those undertaken by precariously employed tech workers (e.g., rideshare, food delivery, and gig workers) tend to be discussed comparatively less often. We further identify common discussion themes, such as (1) complicity in the development of socially detrimental technology, (2) role morality, and (3) attitudes towards unionization.
Background and Methods
The Collective Action in Tech Archive4 documents instances of collective action from workers in the tech industry. For each such action, we matched discussion threads from the social news website Hacker News (HN)5 via the search engine Algolia.6 We did so manually by searching for discussion threads about each of the source URL fields of the collective archive record. Additionally, we made a best effort to search for keywords related to each action to further uncover related discussion threads.
What is Hacker News?
Hacker News is a news aggregator and discussion forum run by the startup accelerator Y Combinator.7 In 2020, it served approximately five million monthly active readers,8 a sizable audience in light of an estimated 4.4 million software engineers in the U.S. (and another 22.5 million abroad).9 HN users post, upvote, downvote, and comment on links across the web, especially those on technical topics and entrepreneurship.
Which kinds of collective action do Hacker News readers devote their attention to?
As a starting point for our inquiry, we sought to determine which collective actions have historically garnered attention on HN. To this end and as a proxy for attention, we ranked actions by the number of comments they received. Various causal mechanisms may explain differences in the number of comments on an action, for example:
Exposure to the Action
Some forms of collective actions garner more news coverage than others. We expected collective actions with more prominent news coverage to be more likely to be read by HN users who then choose to post links on the site.
HN Reader Interest in the Action
Given their technical, economic, and social prominence, large tech companies feature prominently in HN discussions. Some HN readers work at such companies. Hence, we expected collective actions at these firms to be disproportionately discussed and upvoted. HN is an English-language forum. As a result, we expected actions in Anglophone countries, and particularly the U.S., to be disproportionately discussed and upvoted.
HN users tend to be software engineers. We therefore expected collective actions that involve white-collar workers to attract disproportionate discussion compared to collective actions involving other more precariously employed tech workers.10
HN ranks posts based on recency and up/downvotes.11 Being one of the thirty posts on the front page guarantees greater visibility and therefore more comments. Submissions are moderated by a single person12 according to a set of high-level guidelines.13 Politics is considered off-topic, and users have historically flagged posts relating to diversity and inclusion, leading to their downranking.14 Given collective action’s political nature, moderating decisions may have reduced visibility of collective actions on the site (and thus reduced the number of comments on those actions). The platform employs an automatic “flamewar detector,” which downranks threads that attract more comments than votes (unless manually overwritten by the moderation team). These mechanisms are not transparently documented,15 making it difficult to estimate the impact of moderation decisions on overall attention patterns.
The Collective Action in Tech Archive contains 506 actions carried out by tech workers from 1969 to 2022. In the analysis below, we consider only those 477 actions that occurred from 2007 on, as HN went live that year.16 We matched 1,391 HN posts corresponding to these actions, containing 58,984 comments.
Upon inspecting the top ten actions, a few patterns emerge:
All ten actions occurred in the U.S. The first non-U.S. action is ranked #23. It involved Indian blue-collar workers at an iPhone production facility protesting unpaid wages (606 comments).17 This pattern holds true in the overall data, with U.S. actions receiving 1.9x more comments than non-U.S. actions.
Nine actions involved predominantly white-collar workers. Only one action, the Instacart-tipping protest, involved primarily blue-collar workers.18 Overall, actions that do not involve blue-collar/contract/gig workers receive 4.5x as many comments as those that do.
Five actions involved open letters (1.4x as many comments as on actions that did not involve open letters).
Four actions involved Big Tech companies. Overall, there are 1.4x more comments on actions involving Google/Apple/Facebook/Amazon/Microsoft than on actions that do not involve Big Tech.
Two actions involved unionization efforts. This trend does not hold in the overall data, as there are many instances of unionization efforts involving precarious workers that received less attention.
Broadly speaking, we found actions that received no comments to be:
More likely to be carried out by blue-collar workers.
More likely to be carried out by a small number of people in a local context.
More likely to be carried out outside the U.S.
Modeling HN Attention
Combining the above observations, we model the number of comments a given action will receive jointly as a function of the action’s location, participants’ precarity, and the company.
𝒍𝒐𝒈(𝒄𝒐𝒎𝒎𝒆𝒏𝒕𝒔) = 𝜷𝟎 + 𝜷𝟏 (𝒑𝒓𝒆𝒄𝒂𝒓𝒊𝒐𝒖𝒔) + 𝜷𝟐 (𝑼𝑺) + 𝜷𝟑(𝑮𝑨𝑭𝑨𝑴)
The resulting regression coefficients are in line with our univariate insights. Precarity has, all else being equal, a strong negative association with attention on HN. U.S.
location has a slightly positive association. Big Tech has a strongly positive association.
While we were not able to systematically analyze the content of the different discussions, a few themes emerged throughout.
HN readers’ views varied greatly regarding the responsibility to do no harm. This set of differences manifested, for example, in the context of neveragain.tech,19 “an online petition by information technology workers pledging to work against a U.S. government database identifying people by race, religion, or national origin, specifically in response to the Trump presidential campaign[‘s] statements about creating a Muslim registry and deporting millions of illegal immigrants.”20 Amongst other historical references, the petition invoked IBM’s involvement in the Holocaust.
Some commenters reacted with cynicism to the signatories, accusing them of “virtue signaling and keyboard heroism”21 and pointing out the irony that their very profession had recently played a key role in enabling mass surveillance by national governments.22 Others took a stronger view of their responsibilities and expertise:
We're not bystanders; our industry is the most important facilitator of this problem[,] and we understand the causes, implications and solutions far better than anyone. We have a serious responsibility to our fellow citizens, just as the food industry has a responsibility not to poison everyone.23
The pledge inspired another thread in which a tech worker asked their colleagues what type of software they would refuse to develop for ethical reasons, with answers ranging from surveillance tools to weapons to technologies that further addiction.24
To what extent should tech companies be responsive to, and reflect the values of, their employees? This issue may be the most prevalent and consistently controversial discussion topic relating to collective action, going back at least to 2014 when Mozilla employees called on CEO Brendan Eich to resign after it was disclosed that he had “donated $1,000 in 2008 to support California's Proposition 8, legislation that would have made gay marriage illegal in the state.”25
Some HN users argued for a separation of personal and professional spheres:
Personally, I'm so sick of everyone being so damn sensitive about everything nowadays. Guess what? Business is business[,] and 99.99% of the time CEO’s [sic] are appointed on their business acumen and not their personal beliefs. If Eich is a professional he'll keep the two separate.” (emphasis ours)26
In the same vein, other users interpreted Mozilla’s employees’ actions as “bullying,” a “witch hunt,” and downright “fascism.” Other commenters, however, highlighted Mozilla’s stated purpose of “safeguarding and advancing civil rights,”27 which in their view conflicted with Eich’s appointment.
In the wake of the Black Lives Matter movement in 2020, leaders of Coinbase and Basecamp asked employees who they considered to have participated in social activism in the workplace to leave their companies, offering severance packages.28,29 A majority of HN commenters praised CEO Brian Armstrong‘s emphasis on Coinbase’s being “laser focused on achieving its mission”30 instead of becoming “distracted internally by racial justice issues.”31 Others pointed out that “the mission of the company is not disjoint from the society it is embedded in,”32 accusing commenters who were dismissive of employees’ political claims to co-determination of exhibiting a “feudal mentality.”33
HN users’ attitudes towards unionization in the tech sector skewed toward curiousness and favorability, at least in discussions about such efforts at Kickstarter34 and Google.35 Many commenters congratulated those who were involved and highlighted the prospect of addressing power imbalances in the workplace36 and having more of a say in company culture (such as, for example, combating ageism or advocating for equal pay.)37
Considering the professional-managerial nature of software engineering,38 however, some readers questioned whether tech workers were truly workers:
While unionizing for backbreaking work like the manufacturing industry makes sense, in tech it is a nightmare. Let's see where Kickstarter finds itself in the next recession and see how things work out when executives can no longer make quick decisions but are forced to do everything by committee.39
In response, discussants pointed out analogies to other high-earning fields with unions, namely Hollywood actors and professional athletes.40
Much more stands to be gleaned from the HN collective action dataset. Future directions could include:
Using qualitative coding methods and natural language processing methods to arrive at a thicker, more systematic account of the discussion themes.
Making some version of this analysis, perhaps an interactive visualization, available online and sharing it on Hacker News to generate an opportunity for the community to reflect.
Contributing 100+ improvements to the Collective Action in Tech Archive, enhancing data quality and adding additional news sources.
Anonymous was a 2023 FASPE Fellow. They work for a technology firm.
Bao Kham Chau was a 2023 FASPE Law Fellow. He is a visiting fellow at CornellTech, an affiliate at the Harvard Berkman Klein Center for Internet and Society and an intellectual property attorney at Fish & Richardson.
- This author works for a Silicon Valley technology company. They scraped and analyzed the data as well as wrote the paper.
- Contributed to the overall design of the project.
- Inspired by https://observablehq.com/@mbostock/bank-failures
- The Collective Action in Tech Archive allows for each action to be taxonomized by employment type, namely “knowledge workers”, “white collar workers”, “blue collar workers”, “contract workers”, “in-house workers”, “gig workers”, “other”, “open source community”, “interns”, “content creators”, and “retail worker” (allowing more than one employment type per action). For our purposes, we operationalize precarious workers as “blue collar workers, “contract workers”, and “gig workers.”, building on prior work in https://github.com/collective-action/tech/blob/master/analysis/2020-analysis.ipynb.
- Daniel Gackle (dang), who was until 2019 supported by Scott Bell (sctb), see https://news.ycombinator.com/item?id=25055115. For a longer discussion of HN’s moderation, see https://www.newyorker.com/news/letter-from-silicon-valley/the-lonely-work-of-moderating-hacker-news.
- https://github.com/minimaxir/hacker-news-undocumented/blob/master/README.md#implicitdownranking -of-politics
- https://data.collectiveaction.tech/action/62 (6th most discussed action in the dataset)