The Unbiased Hackathon Part 2

This article covers the events during the hackathon, if you would like to read about the planning of the event please read part 1 here.

What is Unbiased and Why?

Unbiased is a software that assists in removing bias from CVs and resumes. It's well-documented that humans are naturally biased. Some people are perfectly aware of their biases and even flaunt them openly, however many of us are completely unaware of biases we may harbor. These biases, unconscious or not, affect the hiring process. There have been multiple studies on hiring biases, a Harvard Business School study found that "whitening" the resumes of African-American and Asian individuals resulted in a 21% increase in interview offers for African-Americans and a 24% increase in interview offers for Asian applicants. Whitening resumes refer to changes such as name, removal of sought-after scholarships or organizations that cater to certain ethnicities, and even changing their hobbies to ones that are more caucasian. Whitening is a way to fit a mold, but it comes at the expense of denying who you are. In the EU we are projected to be 132 years away from gender equality, figures for ethnic diversity are not available for the entirety of the EU. In The Netherlands, it is standard practice (though not forced) to add your photo to your CV. There is no need outside of modeling for a job to care about your physical features.

It should go without saying that your ethnicity, gender, age, or physical features should not be a factor in attaining interviews. However, these are all parts of a standard CV or Resume. Unbiased wants to eliminate all the parts of a CV or Resume that should have no bearing on your ability to do the job you are setting out to attain. These topics and the ideation for Unbiased were discussed in the opening remarks of day one, you can watch them here. The remarks start at the 36-minute mark.

Unbiased MVP and Future

Unbiased started as a seed of an idea, and following the hackathon became a sprout. We at Adabtive have ideas to continue this sprout into a full-grown tree. Imagine the full-service applications we could add to software such as Unbiased! Think about collecting the data on how each hiring manager, department, and company hires. Picture being able to check a report that shows after a first interview that hiring manager "B" is 28% less likely to hire a woman, or that as a company you are 18% more likely to hire a caucasian identifying individual. Unbiased works from the point of view that the company wants equality and that the bias is subconscious. However, these statistics can help identify if certain individuals in the organization are purposefully biased as well. This software can and will grow to the ultimate hiring tool, from candidate matching, and unbiased hiring, to voluntarily collected hiring statistics. Culturally diverse and gender-mainstreamed companies have proven track records for better company culture, higher employee retention, and higher GDP. Diversity contributes to a better product because you have fewer blind spots. Well-intentioned input from a diverse pool is integral to widespread success. You don't know what you don't know, and it doesn't hurt to ask for a differing opinion. Unbiased will continue to grow and morph into what the workforce needs, but here is where we started.

The MVP scope was decided during the hackathon's first day of brainstorming. An integral part of the MVP was breaking down the common sections of a CV, as they are more in-depth than a resume. We realized that almost the entirety of a CV could have a bias in some way. Here are the reasons each section was selected.

• Photo - Physical Features, Gender, Ethnicity, Age
• Name - Gender, Ethnicity
• Email - Often uses someone's name, also allows the user to circumvent the software to contact someone directly.
• Birthdate - Age
• Address - Socio-economic bias, ethnic bias (Think about neighborhoods that are primarily certain demographics.)
• Phone Number - Circumvents the software to contact someone directly, possible ethnic/nationality identifier
• Profile (Profile in a CV is a section that is usually in the third person and tells the story of "who" the applicant is.) - gender, ethnicity
• Career (Similar to profile, but tells the story of "who" the applicant is as a professional.) - gender, ethnicity
• Education - (If year is present)*age, university, or college could give ethnic background. Those who went to prestigious Universities, Ivy League schools can say "Ivy League" over "Harvard"
• Work Experience ( In a CV this is often detailed in the third person) - gender, ethnicity
• Skills - Usually not biased
• Hobbies - Unnecessary for a CV/resume, can give ethnic or gendered information
• Training and Certificates - Usually not biased
• Achievements - Usually not biased

The brainstorming was moderated by our rockstar event mods Heather Johnson, MBA , and Brian Johnson . One of the first things discussed was each section of the CV and then whether or not we wanted to focus on that section or not. Almost all sections needed some sort of change. We discussed what would be ideal and what would be possible during the hackathon itself. Photo was a section that was agreed to be a need to remove, but the how was a serious consideration. The solution we came up with was elegant and simplified the process, but you'll have to try the MVP yourself to see what we did. All participants are students from Codam Coding College , they are brilliant, but many of the technical mountains of the event were new concepts to them. We had lengthy discussions regarding the file format we would accept, the languages we would use, a workaround for storing data without conflicting with the GDPR, and how we wanted to break up each task. The brainstorming itself took hours, from about 9:45 until 13:00. Almost all participants and mentors spoke up and had opinions about the how and why, it was an electric if sometimes chaotic deep dive and resulted in us having a fairly fleshed-out MVP scope.

The MVP works as an email forwarding service that redacts or removes certain information. Nothing private is stored. The hiring manager is considered the client, and they type only the job title. This will be part of a list so imagine this:

1. Full-Stack Dev
2. DevOps
3. Front-End Java

The job title and number will be what they receive as an email subject line when someone applies. What they use to receive resumes is a URL that they can attach to wherever they have posted the job. The URL will take the applicant to a page that looks like this:

The applicant has two options, to upload a PDF or to fill in the form below. In both cases, the applicant needs to add their name for proper redaction, as names could be anything it is more efficient to have them directly inform what their name or nicknames could be and we replace their name with "Applicant." In both versions, the redaction happens after the applicant inputs their information. The applicant does not have to input information that would have been removed entirely if they chose the form instead(I.E we won't ask for birthdate or photo etc.). The information then goes one of two places, if it is a PDF then it is sent to tokenization to parse the CV and remove the information we deemed possibly biased. (Photo, email, birthdate.) After parsing, or if directly from the form it will go to redaction, where changes like name and gender are put into place. Instead of "Candace" it would say "Applicant" and instead of "she" or "her" it will say "they".

Ex. Candace accomplished a 98% occupancy rate. She accomplished this by...

becomes

Applicant accomplished a 98% occupancy rate. They accomplished this by...

This is an amazing step towards equalization. The redaction team also added a library for removing gender from possibly gendered occupations. The software will look for words like "waiter" or "waitress" and replace them with "server".

The next step is sending the email. The remaining unbiased information is added to the body of the email and sent to the hiring manager with the subject being the job title and post number for easier organization.

Once the hiring manager selects a qualified applicant they will mark the email and it will invite the applicant to schedule an interview and give the hiring manager's email at that time.

This is not a perfect solution, but for an MVP it is remarkable and beautiful.

Team Breakdown and Statistics

We had a UX/UI-Front End Team whose mentors were Diego Lopez-Mobilia , Faith Johnson , and Emanuele Parlascino . This team's participants were Emanuela Licameli , Elena Voronin , Dominique Maessen , and others. (Some individuals do not have LinkedIn, or did not give their permission to be tagged.)

We had three back-end teams: tokenization, redaction, and generalized back-end. The mentors of these teams were Chai Stofkoper , Brian Huijbers , Mariah Laqua , and even though he was a moderator he also shined as a mentor, Brian Johnson .

These team participants were Anna Krepkova , Bruno Franco , Daoyi L. , Debora Monfrini , Douwe de Boer, Johanna Guacide , Jeroen van Halderen , Michelle Stegeman , Paris Sadeghi , Sara J. , Yasser Shalash and others. (Some individuals do not have LinkedIn, or did not give their permission to be tagged.)

The remaining team is Testing, whose mentors were Eelco Peeks and Addy Zwiebel . This team was comprised of Zuleykha Pavlichenkova and others. (Some individuals do not have LinkedIn, or did not give their permission to be tagged.)

The other mentors were not specific to teams, but timezone based. This was an excellent choice, often in hackathons mentors are either withdrawn or leave after business hours, giving the participants who work later no mentors to bounce problems or solutions off of. The after-hours mentors are professionals in America, which meant that at the end of our day, it was the beginning of theirs. This was done by the brilliant and lovely Brian Wilkins and Aaron Brandenberger , both of whom provided excellent help and insight. This idea was a success. On the last day of the event, I was met by some enthusiastic participants who worked late, had gotten stuck, and used the after-hours mentors to successfully get unstuck and move forward. I wanted participants who worked hard and late to have resources instead of being frustrated and leaving without a fix.

We started with 27 participants, on the first day we lost 4 participants, on the second day we lost no participants, and on the last day we lost two more participants. This means in total we had 21 people complete this event. From the sign-ups, there were 16 women/women-identifying individuals which means that 59% of the event participants were women! Including the drop-outs the remaining participants were 62% women, an inspiring figure to show how important this idea truly is to women. The drop-outs are also not unexpected, this article from LinkedIn shows that on average 50% of hackathon participants do not finish the event, and we lost only 22% of our participants.

Overview

This was a collaborative event, not a competition, which meant that teams needed to work together, communicate effectively, and share ideas with each other. Communication is a frequent pitfall in tech teams, and in talking to other teams so this was an excellent learning experience(and learning curve) to the event. The hackathon was designed to feel like a start-to-finish MVP creation. We brainstormed, we split off into teams, we ran into issues (think about emails and spam, or redacting properly.), we learned how to effectively work in teams, share when technical ideas needed to pivot, and learned how to integrate a project to reach a functional MVP.

The participants, mentors, and moderators who participated in this event are the best of the best. They altruistically spent three days of their time to create software to equalize the hiring process.

We built a beautiful MVP, this software does good, and now we ask you, reader, will you use Unbiased to assist in your hiring process? If you're interested in trying out the MVP please visit our github. If you want to assist us in making the product better we welcome your help! This is an open-source project. If you want to assist us in building it bigger and better we would love your assistance. Adabtive is working on the next steps regarding Unbiased, but we would love to hear ideas and suggestions from you.

Thank you to the companies that made this hackathon possible: Adabtive , Gozynta , InQuisitive , Blacktree Interactive , and Codam Coding College

* years in your education section are valuable if you are a recent graduate and looking for intern/trainee/junior positions. After this, it is recommended that you remove this information.