Some Employers Use Social Media to Discriminate

Information revealed on social media sites can lead to hiring discrimination, according to research from Carnegie Mellon University.

Thumbs Down from the Left“Our experiment focused on a novel tension: The tension between the law — which, in the United States, protects various types of information, making it risky for certain personal questions to be asked during interviews — and new information technologies, such as online social networks — which (often) make that same information available to strangers, including interviewers and employers,” said Alessandro Acquisti, associate professor of information technology and public policy.

While he and Christina Fong, senior research scientist at CMU’s Dietrich College of Humanities and Social Sciences, found that relatively few employers check up on job candidates online, what they do uncover can influence how they make a hire.

In sending applications to 4,000 real job openings, the researchers compared the interview opportunities of a Christian candidate vs. a Muslim candidate, and the number of interview opportunities a gay candidate received relative to a straight candidate. They also collected data on locations and types of companies advertising the jobs.

“Our survey and field experiments show statistically significant evidence of hiring bias originating from information candidates shared on their online profiles,” Fong said. “Both by itself and controlling for a host of demographic and firm variables, our Muslim candidate was less likely to receive an interview invitation compared to our Christian candidate in more politically conservative states and counties.”

They found less bias based on sexual orientation.

Because the social and political sentiment between areas couldn’t be randomly assigned, the researchers warned the results should be interpreted as correlational, not causal.

Kate Crawford, principal researcher at Microsoft Research, has warned about the potential of hiring discrimination based on the aggregation of publicly available data. She says health information – through health search terms, products researched or bought online — is especially vulnerable to the kinds of subtle discrimination that can result from Big Data analysis. She and a colleague have proposed a framework of “due process” to help consumers understand their legal rights in determining what information about them is collected and how it is used.

10 Responses to “Some Employers Use Social Media to Discriminate”

  1. I applied for a job and the company had the balls to ask for my Facebook ID and Password! I maybe, severally underemployed but I am not going to subject my family/friends to have background and credit checks on them so the employer can use their information against me. This is seriously overstepping the bounds of privacy for me and my friends/family. Discrimination anyone!

    • I’ve heard of companies doing this. It may be helpful to just simply tell them that the Terms and Conditions you agreed to forbid sharing your password. One would hope that the company may just be testing you to see what you do in uncomfortable situations. It is more likely that they want to dig up information, but if you give over the password, you are in violation of a legally binding agreement with Facebook. This way, either the company will give you a higher value because you both pay attention to details and can handle uncomfortable situations ethically and appropriately, or the company is a place you really don’t want to work for anyway.

  2. Google your name before you start your job search. You might be surprised how many People Search websites have harvested personal information about you that an employer should never ask before and during an interview and most questions after your hired because it’s illegal for them to ask.

  3. Experience Developer (4)

    Susan Hall’s post today might lead readers to believe that U.S. companies and conservatives more specifically are discriminating against Muslim and gay job applicants. She reports: “They found less bias based on sexual orientation.” The study says “We find no evidence of discrimination against the gay candidate relative to the straight candidate” but go on to try to prove it a second time.

    The original idea was to create four fake online people with absolutely unique fake names with fake online friends with their own professional profiles and fake colleagues on an unnamed professional network like LinkedIn. You have one gay man, one straight man, a Muslim and a Christian, but this is only revealed on fake Facebook pages. Each fake person applies to over 1,000 real job ads all over the country and the authors count call-backs. Surprise! Fake gay man got 0.01% more callbacks than fake straight man. And there was a statistically-insignificant difference (1.71%) for fake Muslims vs. fake Christians with the full sample, So the authors hunkered down to find the bias they they sought–in the most conservative states and counties in the country.

    The whole exercise is surreal because the authors never prove that anybody actually learns about the applicant’s religion on Facebook. “In an ideal scenario, we would know exactly how many and which employers visited the profiles we designed for the experiment. However, the social network in question does not allow for this kind of tracking.” Not even counts? This from Carnegie Mellon University? Without a leg to stand on, they take 148 Google Keyword Tool searches of those peculiarly unique nondescript names and by page 33, conclude that “…these estimates suggest that a likely proportion of employers who searched for the profiles ranged between roughly one out of ten, and roughly one out of three employers.” This is nonsense to me, taken to four digits of precision!

    Clearly disappointed by the even-handed treatment given to their candidates, the authors dive into the data to find a dismal 2% vs. 17% response difference in the most conservative areas for Muslim guy vs Christian guy. But after reading most of this study, I just don’t trust it. The jobs excluded “…staffing companies, companies located in the same geographic region as the candidates’ reported current location, and companies with 15 or fewer employees.” These conditions just aren’t normal. How did the authors chose their employers? Furthermore, the candidate’s home state is never mentioned in the study (unless I missed it.) Don’t tell me that red states are biased against Muslims when you haven’t controlled for distance-from-the-applicant. Maybe your candidate is from New York and red states just hate New Yorkers because of Wall Street. Maybe the employers in the red states are old fashioned and expect a follow-up call to an electronic job application. Maybe they discriminate.

    But don’t tell me you’re doing a study that shows how Facebook allows employers to discriminate, then accuse people of bias without any proof that anyone, anywhere, ever looked at the Facebook pages that you specifically designed to reveal the religious bias you sought to measure. That’s just plain wrong.

    And Dice is wrong to repeat this without a more careful analysis. Press Release: []
    Draft paper: []

    Please feel free to correct any factual errors I made here.

  4. Awesome response by ‘experience developer’ above. Author apparently didn’t bother to let any of the facts cloud her pre-determined conclusion. This is a non-story, as usual, fabricated by the author to satisfy some pre-conceived notion and all done with non-existent and sketchy facts.

    Thanks Susan Hall for writing a story about some non-existent problem and slapping on a bogus headline in order to attract clicks.


    The headline isn’t bogus–the research paper is “An Experiment in Hiring Discrimination Via Online Social Networks” and it’s a serious study coming from a respected university.

    So Dice techies–if you wanted to count visits to each of four fake Facebook pages passively, could you do it without Facebook’s help? Does anybody know if it’s available with Facebook’s help?


    I would like to retract four elements of my post from EXPERIENCE DEVELOPER (4) of November 30. I hadn’t read the entire paper and was impolite and clearly so aggravated that I couldn’t spell EXPERIENCED.

    (1) “fake, fake, fake…” That was childish. The authors took pains to create unique names and backgrounds, checked for search-engine hits on these names before starting the experiment, and tested the resumes to create believable candidates surrogates.

    (2) “…but go on to try to prove it a second time.” and “So the authors hunkered down to find the bias they they sought–in the most conservative states and counties in the country.” It should be plain to everyone that I cannot read minds and I don’t know that if was post-hoc analysis.

    (3) “Without a leg to stand ..This is nonsense to me, taken to four digits of precision!” I need to clarify that I was referring to the increase in numeric precision from an actual number 3.56% base (3 digits of precision) down to 9.9% (2 digits) and then back up to 27.68% (4 digits), not the increase in the rate itself. It’s a draft. This is the kind of calculation that gets documented step-by-step in a final paper.

    I am so disappointed that the authors could not prove that anyone received the religion-revealing social media exposure. The authors did reference a “…2013 CareerBuilder survey of 2,291 hiring managers and HR managers: 24% claimed to “occasionally” use social media sites to search for candidates’ information, 8% answered “frequently,” and only 5% answered “always.”” That’s very helpful–HR must know this is controversial, so it’s inconceivable that the always-use rate is under 5% for these HR managers. It’s also consistent with the author’s 3.56% Google Keyword Tool figure before adjustments.

    (4) “That’s just plain wrong.” Incorrect. They get to publish and we get to read and decide.

    The authors are accepting comments on their draft. I e-mailed the author and have received some detailed responses to my questions.

    But back to the data–2,087 job applications were sent out and there was a measurable but statistically insignificant difference between pseudo-Christian and pseudo-Muslim applicants. Looking only at the 96 job applications that landed in ten “Republican-leaning” states, nine of 52 pseudo-Christian applications got a response and only one pseudo-Muslim application out of 44 applications got a callback. With an overall response rate 11.78%, you might expect 6.49 C and 5.18 M.

    My gut thinks it isn’t social-media informed discrimination. How many bigots must there be in HR and what percentage of these bigots are checking Facebook pages in order to predict that an expected value of 6.49 callbacks rises to 9 and 5.18 callbacks drops down to 1?

    I’m sending this URL to the authors so they can comment if they’d like. Again, please correct any factual errors you find.

  7. Color Me Skeptical

    “Experienced Developer” (the one who posted 12/2 and 12/3): I hope you will understand that I am having a very difficult time believing one person made such a reversal in the course of a few days.

    For those of us who have developed heightened skepticism after years of reading internet comments, please prove that you are, indeed, the same person who posted as “Experience Developer” and “Experienced Developer”. My understanding is that, on, commenters’ email addresses are tied to names and not published.

    Post as “Experience Developer”, stating that you are the same person posting as “Experienced Developer”. Use the same email address used in previous postings as “Experience Developer”. If, in fact, it’s possible for a Name to used with different email addresses in different comments, I respectfully ask a moderator or admin to confirm the that “Experienced Developer” is the same person who posted on 11/30.

    I’m just trying to protect you from allegations of “astroturfing”. If you are, in fact, the same person, I will not apologize, but I sincerely thank you here for setting the record straight.



    Note that I was the one in post (#5) who objected to someone using the word “bogus”.

    Consider that it doesn’t matter if a stranger contradicts me or I do–the paper is out there for anyone to read. Because it was several dozen pages, I asked people to comment if I missed or misrepresented something. I also sent the URL for this thread to the authors.

    NOTE TO FRUSTRATED IT JOB SEEKERS–in the study, up to two skills or certifications were added to the simulated candidate’s resume if needed to meet the minimum job requirements, and the response rate was still just around 12%! We can ditch any illusions that some rational, benevolent HR person is sitting at their desk with a cup of coffee carefully considering every word on our resumes to find the best candidate for an interview. 12% means 25 applications, some with a juiced resume, to get 3 calls. “If at first you don’t succeed…”