Monthly Archives: June 2015

Getting Only What You Need

Something that has always frustrated me about the Internet is how much data companies require on sign-up and how little privacy there is regarding that data. Name, email, date of birth, country of residence, you can hardly sign up for any service on the Internet without giving some random company the entirety of your life story. This never used to bother me; I used to be a very open person with my information. I was one of those Internet weirdos who, in the web 1.0 era, used her full name.

However, after I came out as transgender my attitude changed. All of a sudden my data was extremely important. Any information I broadcast could be used to target me or identify me as a potential target. This became immediately apparent the first time I was brought into my manager’s office to be told that someone from the Internet had contacted him about me. Whoops, I guess that maybe being open with my data has come to bite me in the ass.

Due to this I’ve become pensive about sign-up forms and online surveys. Why do free and ad supported services want so much of my data? Why would they want to know where I live? I’ve also become increasingly annoyed with the way that companies ask for data. Why is gender choice so limited? Why is race data so Americentric? I’m skeptical of companies that want users to be open with their data, but obfuscate as to why they want my data.

I know I’m not the only individual who harbors these particular concerns either. There have been several high profile examples of data abuse, as well as companies with archaic policies regarding simple collected data. Far too much data is being requested by companies, most of it is bad or unrepresentative data, and most of it is under-utilized. These aren’t difficult problems; they’re solvable if you pay attention to the desires of your users and needs of your advertisers.

Use your words

If you’re designing a tool to collect data in English, lucky you! English is a robust language and there’s always at least a few ways to say the same thing. However, people often forget this and instead rely on the decades old maxim, “Keep it simple stupid,” which is not only ableist but completely wrong-headed in this situation. KISS has its merits in other areas of design but form and survey design, working with data, is not one of these. Lest we forget that some of the simplest designs are the most complex to navigate. Attempting to ask overly simple questions on a form does the same; the problem of bad data doesn’t simply disappear. Use fewer words, but do not boil down complex issues into entry boxes with a single word descriptor, especially if these concepts are defined socially.

If you’re attempting to boil down survey or form answers to one word, you’re essentially subjecting an individual to the textual equivalent of a Rorschach test and trying to pigeon-hole them into an answer defined by the question. “When you look at this amorphous and vague concept, what do you see?” The most important thing is to never forget that race, gender, and, family name, all carry social connotations depending on where your feet are standing in the world. What is “African American” to a “Jamaican Canadian”? What is “gender” to a non-binary individual? What is family name to a Spanish person? Hence if you truly need that information either leave the question more open ended, “What is your cultural background?” or be very specific, “What gender appears on your license?” If these sorts of questions irk you in any way, chances are you have no business asking them in the first place. Surveys looking for census data should leave questions open ended and collate the data later. Forms that have any sort of legal implication should be very specific otherwise you risk having the incorrect information.

Be Transparent

Often I’ll come across a form that gives absolutely no justification as to why they’re asking for information. At it’s purest a sign up form needs two fields: e-mail and password. The email becomes the user id and the password is used to authenticate the user, easy peasy. All information after these two fields is superfluous and often unnecessary. Certainly name and age could be given a pass but most questions beyond that begin to raise red flags as to what a company will be doing with the information. Gender? Why? Will Rdio suddenly block “Ixnay on the Hombre” because I should’ve been listening to “Backstreet’s Back” at the time? Does the Skype app change it’s logo from Blue to Pink? Fuck no.

So why even acquire census information you will never utilize? Certainly, if advertisers are pressuring you, you may have need for this census information. However, if this is the case you’d be better served by stating that plainly to the user. “Yes, this service is funded by ads, that is why your interface will literally be covered in ads, hence we need some demographic information so our advertisers could mis-target you better.” That is all it takes, a small note on the form with an indicator as to which fields will be collated for ad data. Keep in mind the suggestions in the first section about how to phrase those questions but if you must ask them to satisfy your advertisers, do so, and be frank about it, but do not underestimate your user base.

Do not guess

Want to offend someone very quickly? Take wild stabs in the dark at their gender, sexuality, and interests. Come up with an algorithm that uses someone’s speech and topics to attempt to determine their gender and sexual orientation. Make that information available to your advertisers and claim that you have high accuracy when using this algorithm. Sound like a bad idea? Well that is exactly what some services do to get around asking basic census data while catering to advertisers. This is often more offensive than asking questions regarding basic data because it often falls on archaic norms and cultural knowledge than it does on hard data. Especially when built for US-based services, pretending that Silicon Valley’s own cultural knowledge universally applies to anywhere outside of the borders of California, this is a recipe for bad data. California über alles.

Rather than try to be secretive or derive user data through language/topic analysis and divination, it’s far easier to just be upfront with your users. Trying to suss out information via language is Twitter’s answer to “collecting data.” It often misgenders and misidentifies users with almost clock-like precision. From my own experience, there was a month where my gender changed week to week as did the ads. One week it was Tampax and the next it was Glenfiddich. I’m not a genderfluid individual but given the ads that Twitter is feeding me, they have absolutely no idea who I am. Each time it changed, I snapped a screenshot of it and poked fun at Twitter’s poor algorithm. However, the problem is often that topics that aren’t decidedly “female” (e.g., threats to masculinity, feminine hygiene, pink things), are often just labeled as masculine, leading to skewed gender readings. This leaves no middle ground for topics that are gender neutral. All topics must be smooshed into a false binary, and this crap information is fed to advertisers who pay out the ass for this “service.”

This approach, while making the sign up process easier, insults both users and advertisers by providing absolutely unreliable data and feeding that unreliability to the user through mis-targeted ads. It’s often offensive and dehumanizing because it boils people down to data points and tries to rebuild this data into an image of an individual without any concrete vision of the individual. Lastly, it offers the illusion of simplicity while completely and disturbingly overcomplicating the process. It’s an absolutely ridiculous system, especially if this is a pay service that’s being offered. It’s also the absolute nadir of western tech culture, attempting to remove any human interaction and allowing arbitrary data the sole responsibility of determining what a human being is.

Why use a language algorithm when you could simply ask the user what data they’re willing to share to advertisers? Have that information be filled in by the user using optional dialogue on sign up! Is obtaining good census and demographic data really that difficult, given how we have centuries practice in collecting and collating census information? According to the 2011 Canadian Census, 20,535 people in live in the city of Hamilton. 12.8% of the population reported Italian to be their primary language. I could tell you minute data about the city I grew up in, while Twitter cannot tell you your own gender, let alone those of your followers. So tell me, what is the better way to derive information?

Conclusion

Data is what you make of it. Forms and surveys are often one of the few ways you interact with your user base, and the only way you can ascertain reliable demographic information. It may also be the first impression and the start of the trust relationship between the user and the service. Thus a bad sign up form could be ruinous and leave bitterness in the user’s mouth. When you’re going to ask for data remember to in more open-ended ways and do some actual work to collate the data into usable categories. Make sure none of these questions are sensitive or offensive. Make sure they’re not culturally sensitive or ethnocentric. Don’t collect data you will not use. Lastly, don’t underestimate your users or your advertisers. When you can meet those criteria, ask away, I see no problem with these questions provided there is a reason and they are asked tactfully.

If you liked this article, consider reading the related article by Julie Pagano!

On Forms and Personal Information

I am deeply frustrated right now. Yet again, a tech organization has failed horribly at requesting gender information. Their attempts to “fix” it are not an improvement. It’s troubling how frequently this happens. It seems like organizations aren’t learning. This is my attempt at turning my frustration into education.

Disclaimer: I am not an expert on any of this. I strongly recommend consulting people from impacted groups for feedback.

Requesting personal information

Are you an organization that requests personal information from people? Maybe you run a conference and want attendee information. Maybe you develop software and want to learn about your users. Whatever your use case, asking for personal information should be done thoughtfully.

Being thoughtful should include asking a lot of questions:

  • What information do you want?
  • Why do you want this information?
  • What information is required? Why?
  • How will you format requests for information?
  • How will you communicate requests for information?
  • How will requests impact the people you are asking?
  • How will this information be stored? Who will have access to it? Is it secure?

Let’s break each of these down and talk about why they are important.

What information do you want?

Start by making a list of personal information that you want to request from people. You need this for the other questions.

Why do you want this information?

Evaluate why you want each piece of personal information. If you cannot come up with a good reason, you should not be asking for it.

What information is required? Why?

For each piece of personal information, evaluate if it should be required or optional. If you cannot come up with a compelling reason for it to be required, it should be optional.

How will you format requests for information?

For each piece of personal information, consider how you will format a request for that information. Will you give the person a text box or a list of options to choose from? If the latter, how will you populate that list?

Many organizations prefer to use a list of options because it allows for information to be formatted in a uniform way. However, lists are a common source of problems. This post was prompted by an organization failing when requesting gender information by using a list.

An open text box requires more work to aggregate information, but it often provides a better user experience. If you do not have the time to aggregate the information, really think about if you should be requesting it in the first place.

Carina C. Zona’s excellent talk, “Schemas for the Real World,” discusses this and provides some great examples. I recommend giving the talk and slides a look if you want to learn more.

How will you communicate requests for information?

How will you communicate to people why you are requesting this information and what it will be used for? Personal information is important to many people, and they want to understand why you want it and how it will be used. Thoughtful communication not only informs people, but can assuage worries about privacy or even make them feel good about your organization.

How will requests impact the people you are asking?

Think about how your requests will impact people. Consult with people from impacted groups to test things out and get feedback. You should care about this. It is a part of the user experience for your event, product, etc.

Some requests can lead to a negative response. Will requiring private personal information make them less likely to use your product? Will a poorly constructed list make someone feel unwelcome at your event?

Some requests can lead to a positive response. Will asking people about their dietary needs make them feel welcome at your event? Will providing an open text box for personal information make people excited about your product because they can express themselves accurately?

How will this information be stored? Who will have access to it? Is it secure?

For each piece of of personal information, evaluate if it is personally identifiable information (PII). PII has security (and sometimes legal) implications. I am not an information security expert or a lawyer, so I am not qualified to expand on these implications. If you are asking people for PII, you should consult with experts about these issues.

Gender

Gender is the piece of personal information I most frequently see organizations get wrong. One such failure prompted this post, so I want to expand on this.

Many organizations request gender information when it is not needed. Sometimes out of laziness. Sometimes out of habit. Really think about why you are requesting gender information. If you don’t have a compelling reason, don’t ask for it.

Assuming you have a legitimate reason to request gender information, you need to be thoughtful about how you do so.

Let’s start at the beginning. Gender is not binary. If you give two options: male and female, you are doing it wrong. Let me repeat. GENDER IS NOT BINARY.

Some organizations want to indicate they are inclusive of transgender people and try to do so in their gender form. This often goes terribly wrong. For example, look at Grace Hopper Celebration 2015’s registration form below.

on-forms-and-personal-information--ghc-reg01
Original gender input for Grace Hopper Celebration’s registration.

They get a point for understanding that gender is not binary, but fail pretty badly by including “MTF” (male to female) and “FTM” (female to male) in a narrow set of options. While some transgender people prefer to identify this way, many would find this a frustrating and upsetting choice to make. For example, a transgender woman is a woman and would likely want to select the female option in the dropdown.

After receiving criticism about this, they modified the form as seen below.

on-forms-and-personal-information--ghc-reg02
Modified gender input for Grace Hopper Celebration’s registration.

This version of the form is about as bad as the previous one. While some transgender people may be ok with this, many are not and find it problematic. This form makes them choose between their gender (e.g. female for a transgender woman) and sharing the information that they are transgender. Most people use transgender as an adjective or modifier, not as their gender. I am a cisgender woman. Some of my friends are transgender women. We are all women working in tech, and we find this gender dropdown unacceptable.

Both iterations of the form can make people feel obligated to disclose that they are transgender. For many, this is information that they do not want to disclose for a wide variety of reasons including privacy and safety. This requirement may upset people and make them feel unwelcome or unsafe at an event. The organization’s interest in obtaining this demographic information should not trump the comfort and safety of potential attendees.

How do you handle the complexities of gender and make sure you do the right thing? Use a text box! Below is an example of the gender input on the registration form for Open Source Bridge 2015. They also make it optional, so that people are not obligated to share this private information.

on-forms-and-personal-information--osbridge-reg01
Gender input for Open Source Bridge’s registration.

Why are you picking on Grace Hopper Celebration?

Here I am, yet again, writing a blog post about how Grace Hopper Celebration has failed. I’m not happy about it, but they leave me little choice. I expect more from an event that bills itself as “the World’s Largest Gathering of Women Technologists.” An event run by the Anita Borg Institute, a prominent organization with a budget of several million dollars. I am not picking on them. I am trying to hold them accountable. As a member of the demographic they supposedly serve, I expect more.

As a note, they published an explanation and apology about what happened, but I find it lacking. I expect more.

We recognize that a write-in box for gender identity that allows participants to simply mark their own gender identity is the optimal solution, but are currently unable to provide this service given our current resources.

Conclusion

Please be thoughtful when your organization requests personal information from people. Ask a lot of questions. Only require information that is critical. Be thoughtful about how you ask for information. Communicate why you want the information. Empathize with the people you are asking. Get feedback before you move forward on a path that may end badly. Make sure you have the resources available to address problems and make changes in the case that you make a mistake.

On Open Companies, Consent, and Safety (among other things)

This article originally appeared in Model View Culture.

An audio version of this article will be made available on SoundCloud shortly. You can follow Inatri on SoundCloud to be notified when the audio version of this article and others are made available.

There are two goals of Open Companies, as I understand them. The first is to create companies that are actually considered trustworthy, instead of barely above the legal minimum for trustworthiness. That is to say, companies that go out of their way to make sure that they are doing right by everybody who interacts with them – owners, employees, customers and/or users, society at large, etc. The second is to create an environment where people can more easily become involved with the company.

While I like the ideals and goals of Open Companies, I find that the approach tends to position openness as an end in and of itself, instead of as a means to reach higher goals and values. While well intentioned, this approach to “openness” creates a shallow show with minimal tangible improvement, while also strongly reinforcing the problematic under-representation of marginalized groups.

I have discussed Open Companies before, including in Model View Culture’s previous interview with me about my work as a User Advocate for Gittip. The main purpose of this role was to bridge what appeared to be a communication gap between users and the primary development team. I ensured user feedback got to the appropriate places it needed to go in order for it to be acted upon, by breaking discussions down into actionable items. I also provided a means for users to provide feedback in a way that was entirely anonymous to the rest of the Gittip team.

The general idea of Open Companies, as proposed by Chad Whitacre (the founder of Gittip) and elaborated on later, both in person and by the Open Company Initiative, is fascinating to me. I call out Gittip quite often throughout this article, as a side effect of it being the only Open Company which I have been closely involved with. I am doing this in an attempt to both help Gittip improve in these areas, as well as help other companies avoid these issues to start with.

Recentering: Outcomes over mechanisms

The vast majority of the issues Gittip has experienced when interacting with other people, groups, and companies seem to arise not necessarily from a difference in ideals and values, but from a difference in perceived prioritization of these ideals and values. By treating openness and transparency as an end in and of itself, instead of a means to an end, it gives the impression that openness and transparency are prioritized over consent, safety, and comfort — regardless of whether or not this is actually the case. In our attempts to create trustworthy companies, we have unintentionally made the situation worse: like traditional companies, our ignorance tramples those that wish to help us, but now we have turned it into a sideshow for all the world to see.

As we move forward, we need to treat openness and transparency as tools, not the end result. Let’s recenter on the problems we’re trying to solve—instead of the tools being used to reach it—and build out from there. Let’s stop pretending we can use one simple solution for a multitude of challenging problems.

Let’s give safety and consent the absolute highest priority, with openness and transparency prioritized explicitly below those. This means digging deep, properly articulating in detail what problems you are trying to solve with openness and transparency, and handling them individually or in smaller groups.

This will likely prove to be difficult, and needs to become an ongoing process, instead of a one-time occurrence.

The problems that need to be solved will also vary from company to company. However, the end result of this effort will be a more trustworthy company than the current approach can accomplish — and, at the highest level, this is exactly what we’re aiming for.

Transparency: Behold, the cinderblock

It is very easy to get from the ideal of Open Companies to the implementation that Gittip has taken: dump every bit of data you have about the company into a public medium. However, what you wind up with is a company that produces so much unorganized, uninteresting and irrelevant data that you can’t find meaningful information.

You try to find out what has been worked on over the past week, but it’s buried in the midst of an hour and a half of recorded, “open” videos. Even though it can be properly explained in a single-page summary, it’s left only in a time-consuming format. I know of absolutely nobody who has put forth the effort to translate Gittip’s recorded calls into a more usable format on more than one occasion.

“Transparency” by this definition is meant to help people see, in detail, inside of the company… even if it’s wholly uninteresting to the vast majority of people.

The thing is, this approach is not transparency, it’s a fucking charade.

A far more approachable alternative is to find out what information is wanted, and by whom, and break it down in such a way that they can get the _information they need_ without digging through wholly irrelevant noise. Organize it in such a way that it can be accessed, searched, and cited, easily. As an example, I’ll approach the issue of chatrooms, since it is one I have a lot of experience with. The most successful approach I have seen is having a public chatroom with a few people who have access to the logs, and having them go through and create minutes of meetings and important conversations. The minutes are then made publicly available.

This can be done regardless of any company decisions about the publicity of internal discussions, and thus is applicable both to companies like Gittip, as well as companies that are more reserved about publicly logging all discussions.

Openness: Your unlocked door weighs more than I do

My understanding of the term “openness,” as used for Open Companies, is that anybody who wishes to participate should be able to with minimal friction. The inner operations of the company being transparent and easily understood assists with this. The idea is that if everything that can be public is made public, then the onboarding process is just formalities and handling the legal side of things. However, the approach that is chosen is often what I feel can accurately be described as “radical transparency” — the very problem I pointed out in the previous section.

This approach is highly problematic for marginalized groups. One of the things I keep seeing pushed is publicizing people’s salaries. In our world, where marginalized people get harassed over the amount of money they make, it is sometimes not safe for a person to have things such a their salary made public. Some people (including myself, previously) have suggested simply making it opt-out. However, if you make it opt-out, it doesn’t solve the problem: the people who have opted out can very easily be inferred by omission.

Instead of trying to make a one-size-fits-all solution, we need to step back and solve each openness-related problem on its own accord.

Publishing all of the data you can doesn’t really solve any problems. For salaries in particular, I’d argue that, really, nobody cares what other people’s salaries are unless the ratios of compensation to cost of living are unfair. If two people are considered roughly equally valuable to the company, they should be able to afford roughly the same things. This isn’t about exact dollar amounts, it’s about the ratio of cost of living to compensation, and it is really fucking hard to get right.

So how do we solve this? One idea is to discard the self-selected salary approach that has become popular at some startups, and instead set clear terms for specifying how much somebody is compensated. This would likely be based on experience, how long they’ve worked at the company, and the cost of living in the location they are in, among many other factors.

The key is to make this process well documented, as objective as possible, and everything about it publicly documented.

If an employee feels they are not being fairly compensated, they can check for themselves that they are being compensated as much as the process specifies they should be. If they are not, they have what they need to speak out about it. If they are being compensated as the process specifies, but feel it is unfair, they have what they need to present a case for process being adjusted.

The key is to look at the specific structures and ramifications of a problem, and coming up with the solution that makes the most sense, rather than simply defaulting to an arbitrary notion of openness that may create structural problems of its own.

Safety: Implicit is insufficient

One of the things that has come up multiple times with Gittip is personal safety: the ability of someone to participate in the company’s activities without feeling they are making themselves vulnerable or putting themselves directly in danger. A prominent example, mentioned in a GitHub issue for Gittip back in December 2013, is that some people get harassed simply for asking for support for their work — including being accused of “begging” and people trying to police their spending. There are also many who have voiced weariness over joining publicly-recorded conversations, for reasons related to harassment (harassers having the ability to see what they say without anyone knowing of their presence) and privacy in general. The only way to handle this sort of problem properly is by explicitly placing consent and safety over openness and transparency.

I previously proposed that all companies should be explicit safe spaces. The problem is that there are varying definitions of safe spaces, and some people claim you can’t have somewhere be a safe space for everyone. I feel a space can be made safe for every person, but not for every idea.

Simply claiming that a company is a safe space is not enough. You must state who it is intended to be safe for and a non-exhaustive list of which ideas are explicitly allowed and explicitly disallowed. These lists always exist at least unofficially.

It’s time to Fucking. Own. Them.

Chugga chugga *spaceship noises*

At the end of the day, Open Companies don’t sound like a terrible idea. In fact, the end goals they’re trying to achieve largely overlap with my own: creating companies that are run in a way that makes them as trustworthy as possible, and making it as easy as possible for people to become involved. The problem is that the approach of publicly releasing all of the data you can is a non-solution: it is an approach that makes it _appear_ that you’re taking action, but in the end it winds up creating an echo chamber for the existing problems in the world.

To me, the reason it turns into this echo chamber strongly mirrors what Andi McClure has previously said about college admissions processes: when you start with a system which is blind to race, class, gender, and disability (among other things), and throw it into a society which is very much not blind to them, what you get is (at best) a repeater of the previous and ongoing discriminatory acts that society as a whole perpetuates.

I got excited and jumped on the Open Company train. It seemed like such a magnificent idea! The problem is that it is not a train, but some glorious creation from a time where power dynamics and discrimination are almost wholly irrelevant.

Your train is a spaceship from the future, and nobody knows how to drive it. When laying out how a company is to be run, you absolutely have to take into account the realities of the world you live in. The reality of our world, right now, is that the simplest approaches to openness and transparency wind up simply creating yet another place where marginalized groups lose their voices.

These problems are hard, possibly even impossible, to solve entirely. However, we can do better. We can do far better. This is me calling on ALL companies not just Open Companies — to step up. It’s time to stop sweeping these problems under the rug or trying to find a one-size-fits-all solution, and tackle each of them individually.

We can’t get to the future we want by pretending we’re already there.