Thoughtworks accelerator programme: The Digital Union

Back in November, I spent an interesting day in the company of dozens of Thoughtworks staff and others from London’s tech community. 

I was one of 6 presenters talking about real world problems that were affected by technology, and where a different use of technology could potentially make a large difference towards a more equitable society and economy.

The problem I wanted to share was that work is becoming more flexible, but that the benefits of this flexibility are often only going to the employer, not workers as well. Technologies that could give workers more control for a personal relationship with their work have instead been used to shift business risks from employer to worker. And as work becomes more “flexible” and managed by technology rather than people, workers’ ability to exercise power through collective action gets ever more eroded. (read a bit more on the problem here)

I was interested to hear what solutions Thoughtworks’ new Equitable Tech Future community might come up with to this problem. It’s an ideation space that Thoughtworks have convened for their own people and the wider tech community, as part of the strong social pillar to their business.

We discussed issues around co-design of work systems to make flexibility genuinely two-way, and how the digital principles of transparency and collaboration might help workers to find a new type of collective to challenge increasingly unbalanced power relations at work.

Ideation

We broke into groups for the different problems, and six of us chose to focus on the flexible work question. With two Thoughtworks staff facilitating, we started with a visioning exercise, imagining what the experience of workers would be in a world 5 years out, when this problem had been dealt with.

 

We wrote and grouped a large number of answers, and came out with three broad groups around raised awareness of rights at work, greater transparency of previously hidden information, and technological spaces in which workers could readily come together digitally and form networks themselves, within employers or wider industries.

Then mapping back through the years, we decided on milestone interventions that could move us towards that vision, ending up with a rough plan for a multi-stage project that united these three areas.

The Digital Union

We called our project The Digital Union. It would be a social network and organising space, where workers could comment on their jobs, read each others’ comments, access related advice sources. Behind the scenes, the system would start to build up data repositories and networks that could help them do something about it.

Users would contribute a short summary of their problem, which would build into profiles of companies or industries. They could see and comment on others’ related experiences, building a sense of collective problem, rather than isolated individual cases.

The site would help them to access statutory or voluntary enforcement routes and advice sources, in the cases those would be helpful, via an automated triaging process.

As the User Generated Content grew, networks could form, communicating news and developments to people with similar preferences – sometimes automated, sometimes from active moderators. Linking in to unions, NGOs and social networks, would help workers to take up their issues themselves in campaigns or local organisation – where (as in the majority of cases), enforcing rights at work was not feasible for that worker alone.

The Pitch

Next we worked on our elevator pitch, refining it with a simple template:

The Digital Union is a product for all workers who have experienced problems with their employer. It is an information sharing platform that challenges employers’ advantage of control of information. Unlike Facebook and Glassdoor, it is geared towards tackling those issues through practical actions.

And we took our work to the wider group to pitch it.

Acceleration

We were surprised but very pleased to make the cut when the votes were cast for projects presented, and we started a series of 6 fortnightly 2 hour meetups, where we worked further on aspects of the project.

More Thoughtworkers joined us over the time, and we had a number of others from the wider community drop in and out of the the group from week to week, bringing with them new perspectives and different expertise, from development through UX design, legal and copywriting.

Defining the project

We decided early on that the project from the ideation day was very broad, and we needed to focus specifically on narrower aspects of it.

Possible functional areas broken down were:

  • Forums and reporting issues – How we could make this safe and easy to use.
  • Aggregating data to spot patterns that could be useful to workers or organisers.
  • Linking people together with networking or organising tools.
  • Triaging people’s problems for practical action

We chose to use the time in the Accelerator Programme primarily to demonstrate the idea of using work-based commenting to show potential for building awareness and networks.

This outline goes rather further than an MVP (!), but covers some of the more detailed ideas we had on it.

Gathering and aggregating stories

Users are asked to leave a story about an issue at their workplace, for others to read.

  • Stories can be viewed by company / issue / industry.
  • Stories can be upvoted by other users and/or commented.
  • Would adding company be optional if people were scared? Eg industry and issue only?
  • Do we want other filters, eg region/nation?

Privacy & anonymity

It’s a hard ask to get people to part with contact information, when they may be concerned about reprisals or tracing. But conversely as work problems are not necessarily a key part of someone’s identity, they are unlikely to revisit without a prompt, and registration to allow push notification would be key to later stages of the project.

The flip side to the freedom of total anonymity if you can’t show the persistent existence of different users, you lose a large degree of social proof.

Others may not realise the significance of what they’re writing and get themselves into trouble by making themselves identifiable. Our project would need to establish its value to the user before asking for contact information.

  • Enforce a username format (eg funny randomised format [Title] [Adjective] [noun]). Users can roll another one, but not change it themselves, to avoid accidentally creating a profile that gave them away. We’d explain that this is to keep things anonymous. Making privacy into a kind of game at this point could set the tone, and encourage people to think about what they’ve written in their comment as well.
  • Don’t ask for contact details until after people have had an experience of voting/commenting. And make clear that contact details will be to send them updates on whether people like their idea, or more in areas they work, not to market to them, or share with their employer. We considered more techie-friendly privacy solutions such as stored codes to enable contactless registration, but the complexity was too great for the majority.
  • Consider potential for an algorithm to clean up/flag comments for safety? Eg “it looks like you’ve named your location / typed a colleague’s name, do you really want to do that?”

Finding users

Most people in insecure work wouldn’t ever search for information about it. Those that do are likely to be aware they have a tangible problem. How would we go about finding an audience for this project?

  • Initially use TUC advice website (worksmart.org.uk) to advertise comments on relevant issues from users, drawing them into looking up their own company.
  • Build up public issue/industry/company pages that will be a good trawl net of content on search engines.
  • Aggregated data and stories could be used to generate PR to promote the site – eg “non-payment of wages is the biggest issue in hospitality finds new Digital Union”
  • We will rely mostly on user-get-user for cluster building. We should offer private share options (eg Messenger, WhatsApp) rather than public (Facebook, Twitter) to get people to send a story secretly to a friend on mobile.
  • Where clusters are starting we could conceivably look at PPC search and social ad targeting to find more people at a given employer and build critical mass.

User journey

The site should be mobile first. People in insecure employment especially will have smartphones, but might have less regular PC access.

Users won’t think to check the site more than once unless they’re very committed. How can we find a way to develop continuity over time and work toward a network where clusters appear?

  • Email updates.
    • Automatically serviced eg monthly “Your issue has received 2 new votes and a comment”
    • Also include popular new issues added in user’s company or industry.
    • Could include short news content added by moderator and sent to relevant users. eg “Have you see what happened at X Coffee?” to fast food sector

Networking potential

As the constituency for this site won’t all be up for networking, they’ll need coaxing and an active moderator would need to encourage community. Yet finding the resource to moderate this would be hard if it weren’t bringing in its own revenue.

Steven highlighted a good example in Polis [link]. It’s a way that issues can be aggregated and through voting and commenting, the most popular can rise to the top. This way some of the ‘business intelligence’ functions could be either automated or crowdsourced.

  • Moderator/organiser could identify bubbling up groups and use it to identify grassroots leader candidates where unions could be identified to support them. Eg “We could present this to your employer, with the shopworkers’ union USDAW. Would you be willing to get involved?”
  • The site could be used to provide aspects of social proof during existing union organising campaigns.

Deepening our understanding

We split into interest groups to deal with three areas where we thought we could usefully expand our understanding of the project: Privacy and anonymity, triaging advice questions, and simplifying the user experience. EG: notes from workshop on questions around privacy and anonymity.

Building towards a demonstration

After this, it was on to developing something more tangible. We had found a number of questions and potential solutions that could be helpful to many people working in this space, particularly in unions, and wanted to be able to help people grasp ideas more quickly with tangible demonstrations.

Rachael hit on a good solution for the issue of triaging employment rights advice  in a more narrative form. Rights often are less useful if you don’t know them or can’t afford to enforce them. Often a step towards organising at work is more powerful in dealing with a particular situation than a step down a legal route that may go nowhere.

By building questions into decision trees rather than text information, we can have more of a personal feeling approach to helping people find advice, and one that doesn’t demoralise them with the whole huge range of less relevant options. For example, a problem may have a range of different solutions that could be around statutory enforcement, legal rights, personal relationships at work, or collective action and campaigning.

We investigated a simple tool for prototyping – the interactive story format Twine. Joe and I looked at this, and found it would indeed serve the purpose – though the level of understanding required to make more complex branching questions and variables was too great for the time available on this project.

Tom, with Shenday and Karrom, worked on visualisations for the user journey, helping to demonstrate how to take people through complicated actions in an intuitive way on a mobile screen.

After presenting and refining these with the group, Tom then developed them further with prototyping tool Figma.

Steve and Joe then started coding. They made a basic prototype username generator and comment adding system, following the group’s mobile-first design steer.

The code for this has been released on GitHub for others to download and check out or adapt (get it here).

Results

We presented our work back to Thoughtworks’ Equitable Tech Community one evening in late January. It was great to see progress the other strand of the accelerator had made too (Humanity X seeking to increase transparency for transactions in international charity donations).

The project is out there on GitHub now, and using Slack, the group are keeping in touch with options to see where the ideas from the project might develop. If anyone wants to learn more about it, please do get in touch.

I’d like to thank everyone who organised and contributed to this program, but special thanks to Charlotte Fereday at Thoughtworks who organised the program, and our fantastic Thoughtworks faciliators Steven Liu and Joe Wroe. Tom Byrne contributed a huge amount to the vision and execution of the project too, and it wouldn’t have got nearly as far without his work and ideas.

For me, a hugely valuable part of the process was the time we were able to spend with experts in different disciplines, working on issues that we needed to crack if the project would be realistic. These are issues that are going to affect unions and others working in this area, and the learning from it will serve well in many projects in the future. I’ve grouped this into 5 areas that I hope others will find as useful as I have.

Digital union: 5 questions that we didn’t know we had when we started this

1. How can we lower the barriers to participation?

Rationale:

  • We’re asking people to register for a user account, and trust us with an email address and personal information. That’s a big disincentive to get involved with anything.  
  • People in any case tend to believe doing something about their situation at work is pretty futile, so this has to tap into being fun, having a theory of change, and being extremely easy.
  • Getting accurate personal information is important, how do we gather that whilst not making users’ interaction onerous?
  • We need to protect users from trolling. That has the potential to kill the site if people see their stories will only be met with indifference or abuse.

Suggested solution:

  • We could maybe start people off with easy asks like +1ing other stories, before asking them to register or add their own stories.
  • Can we focus replies to maybe emoticons rather than comments? (eg “me too” or “hugs”). What could different easy tags tell organisers about how widely felt an issue is?

2. How can we develop a mostly automated content journey to keep people engaged and deepen their involvement?

Rationale:

  • Matching groups of users together only works if we have strong contact routes to get back to people, and people remember they signed up to the service in the first place (eg they don’t just get a call two years later when a match shows up). Most users won’t think to check the site again later, and even if they do it might not have new content for them. So push as well as pull will be very important here.
  • Getting accurate personal information is important if we’re going to be showing them content that is relevant enough to them to keep them subscribed.
  • Getting back to people that others like or comment on their stories will keep them more committed.
  • Showing people regularly that others have the same issue will help people realise that their problems are large systemic ones as well as being ones that affect them personally, and help them make the connection that they might be solved collectively.
  • People have extremely low expectations of what they can expect from work. The chance to add in content about successes and news stories where things are better would help get people thinking about the possibility of change.

Suggested solution:

  • We generate emails that are mostly based on UGC matching personal preferences. Eg “Someone voted for your story” or “Read three new stories from users at Costa Coffee”. This keeps people coming back to the site, and keeps them thinking about it and boosting new users’ involvement. Most people will probably only get this stuff mainly.
  • We have a way for moderators to insert relevant content into the system by company or sector. This could include news stories, campaigns at a particular employer, surveys (eg surveymonkey links) or paragraphs to update them on the results of surveys. This is how we can mobilise people, educate them more widely, or seek to contact them. This will mostly be content seen by people in companies that organisers are actively working on, rather than general user base.
  • Do we consider a two way channel to help moderators respond to comments from emails, and allow them to answer the most questions most efficiently? (eg Relay)

3. How can a moderator/organiser spot useful patterns in the data and work to link up users?

Rationale:

  • We’d like to use this tool as a way to help workers organise to sort out their own problems, by joining unions or just by working together more. The key to this is finding a way to link workers together.
  • There will be different options to doing this:
    • Via unions. If a moderator sees a group of people in the same employer, they could contact the union to see if they were interested in offering support. Unions are generally not able to support potential members if they’re bringing an expensive problem to sort out, whilst no guarantee of income from membership. However, where a group is forming and a union can help them to do more themselves, that changes things. A moderator could identify a group within a workplace, and offer to them or at least the more active people from it in touch with a relevant union, and set up contact with an organiser – eg a chat session.
    • Via campaigning organisations. The Organise Network have recently started working on employee-started online campaigns (echoing work done by Coworker.org in the States). A moderator could spot an issue blowing up and find a partner group who could support workers to run a petition.
  • To do this, the moderator will need to have a dashboard that shows them the state of the database to date. Eg a table of companies that have become more active recently, and top issues within those companies. What else would be helpful to know on the dashboard?
  • The moderator would also need to be able to get reports on particular issues, companies and industries, with numbers against them. This might help get the evidence to make partnerships with other organisations.
  • Some users may be particularly active and engaged. These people might be good grassroots leaders to work with and to help catalyse a less active group

Suggested solution:

  • Back end of the website will flag up cases where groups are starting to form, spotting trends by company, sector or industry that an organiser might not otherwise be checking for. This will help organisers to spot opportunities to help earlier on, and work on campaigns as they become topical.
  • It will let moderators search to find and categorise emergent groups for later use. Groups will then update based on new additions to those categories, and the number will be updated in the CMS.

4. How can we allow for persistently displayed usernames whilst still ensuring privacy?

Rationale:

  • Showing a persistent username across comments and stories helps show others that it’s a real person interacting with them, and increases sense of trust and community with other workers.
  • Lack of trust in their colleagues is one of the key barriers to workers considering collective solutions to their problems at work.
  • Using real names is out as we need to show people they are unlikely to suffer repercussions at work from speaking out on a website like this. Need to anonymise the usernames.
  • Users may subconsciously give away information about themselves even if choosing an anonymous nickname. They could pick something they use elsewhere, or a name that gives away enough information for someone to guess.

Suggested solution:

  • We give people a choice of nickname only from randomly generated ones.
  • Nicknames could be a combination of titles, nouns and adjectives, with option to spin again and get another totally random suggestion, but not to edit.
  • Three parts to the name could make for a large number of potential names from a
  • Funny names and odd ones would make it a fun feature that makes the service more appealing to use – gamifying user registration eg should you take the current suggestion or gamble on a better one?
  • Done well and explained well, this could have the beneficial side-effects of making people think more about their security in any comment they’re writing.
  • Do we extend this to a randomised avatar as well (eg gravatar) to make content more visually appealing?

5. How can we use user generated content to help us find people online?

Rationale:

  • A lot of traffic to this will have to come through search. People are less likely to share their own story if it makes it look like it’s theirs or if they think it could get them into trouble, so social is out for most people.
  • We will be publishing lots of content about companies, issues and industries. How do we display that content such that it will do well in search engines?
  • We have to find a way to let users categorise content that will be transparent to people. Using straight SIC & SOC codes for example is complicated & will reduce completions. But if we use a tagging system we could end up with too many different tags, and content that could be linked together would not get enough results to make a decent page.
  • It’s very hard to talk about employers in a way that makes sense to people – with agency subcontracting, holding companies and subsidiaries getting in the way. Could we do something like matching workplace data from Google maps instead? Could this become a learning system, able to help decide which sectors companies work in, based on initial links that people are making?

Suggested solution:

  • Display lists of content by industry, employer, issue. Surround them with content that makes it likely it will do well in search.
  • Find a way to integrate google maps data or other source into identifying employers.