Why is racial discrimination illegal on employment websites but not on social websites such as Facebook or Tinder?

Junfang Zhang, Hwa Chong Institution (College), Singapore

Second Place for the 2020 Law Prize ​| 8.5 min read 


The hiring of new employees, and the selection of content to bump up on social websites: both are underpinned by the goal of establishing a “best fit” — between employer and employee, and user and content/profiles suggested respectively. Why is racial discrimination prohibited in the former, while race-based filters and matching algorithms remain common practice in the latter?


It is almost unequivocal today that distinction with regards to race in employment should be condemned. Acts such as the Equality Act in the United Kingdom and Civil Rights Act in the United States disallow employers from requiring racial details that do not relate to one’s ability to perform the role, or to implement policies that disadvantage a certain race. The principle that underpins this is equal opportunities — that only qualifications, ability to do the job and relevant experience should be considered in selecting job candidates (Boyle et al., 2005).


Yet, a very different standard seems to apply on social websites. Of the 25 highest grossing dating apps in the US, 19 requested users to input their race, 11 collected users’ preferred ethnicity in a potential partner, and 17 allowed users to filter others by ethnicity (McMullan, 2019). TikTok uses an algorithm that mimics the physical characteristics of a user’s profile picture in subsequent recommendations, including the colour of one’s hair and skin (Gassam, 2020). Though TikTok’s spokesperson expressed that recommendations are “based on user behavior”, the point remains that racial features are considered to affect the “likeability” of a profile.

In this essay, racial discrimination on employment websites shall be scoped to discriminatory practices during hiring, where race influences the candidate’s job suitability. Racial discrimination on social websites shall refer to acts of collecting and utilising racial information, whether by the user or platform, to create the best match between user preferences and content/profiles suggested. Though other forms of racial discrimination on social websites exist — famously Facebook’s alleged race-based advertisement targeting — the definition above provides the closest parallel to how racial information may be utilised in employment, from which more meaningful comparisons can be made.


This essay shall explain the different standards regarding racial discrimination by making the case for employment and social websites separately.


Racial discrimination and employment


The arguments for why racial discrimination is not justified in employment are as such:


  1. Hiring serves to find the most suitable person to perform a particular job; given that empirical evidence shows that race does not influence an individual's work ability, it is unjustified to allow racial information to sway hiring decisions.

  2. Employment is inextricably tied to standard of living and quality of life. If we accept that everyone has the right to a standard of living adequate for a dignified life, no person should be disadvantaged in that pursuit on the basis of race.


Let us first contend with argument (1). The world of work and recruitment is as old as civilisation itself. In ancient China, imperial examinations served to select officials for the state bureaucracy based on merit (The Editors of Encyclopaedia Britannica, 2019); the Spartans of ancient Greece were notoriously selective in which young men could enroll in their elite army. Times and contexts have changed, but the rationale underpinning these selection practices have not: to hire personnel with appropriate skills and abilities to meet the needs and requirements of organizations (Kapur, 2018).

That established, there is a plethora of scientific evidence that shows race does not genetically determine intelligence or capabilities. Following the controversial publication The Bell Curve by Richard Herrnstein and Charles Murray in 1994 which found differential IQ scores for different races, the American Psychological Association established an 11-person Task Force to evaluate the book’s conclusions. With regard to the cause of the mean Black–White IQ score difference, the Task Force concluded: “There is certainly no support for a genetic interpretation” (Neisser et al., 1996). In fact, as suggested in 2012 study by Hampshire et. al further, differences in intelligence test scores are driven by “other correlated demographic variables such as socioeconomic status, education level, and motivation.” (Hampshire et. al, 2012) Since race has not been found to influence an individual's capabilities, it is unjustified to allow racial information to sway hiring decisions.


It may be objected that a discussion of social issues addresses cognitive and biological studies to such lengths. However, it is important to recognise that as unthinkable as “biological racism” might be to our modern mind, theories concerning racial differences in intelligence are age-old and antedate empirical studies by thousands of years (Eysenck, 1984). Greek and Roman writers in the centuries preceding and following the birth of Christ had much to say about the weak intellects of “barbarians.” Carolus Linnaeus in his Systema Naturae (1758) ranked the various races by appearance, temperament and intelligence, putting the European man at the top, and the African man invariably at the bottom (‘crafty, slow, foolish’). Thus, it is important to firmly establish that there is no empirical basis for differential performance due to race.

Let us now proceed to argument (2). Standard of living can be defined as “the aspirations of an individual or group for goods and services” (The Editors of Encyclopaedia Britannica, 2018) It is undeniable that employment is inextricably linked to standard of living. The outcome document of the United Nations Conference on Sustainable Development recognises in paragraphs 143-157 the linkages among poverty eradication, full and productive employment and decent work for all (United Nations, 2012). The kind of employment opportunities we get (or, as above, the availability of employment in the first place) directly influences the income we receive. This in turn determines the amount of goods and services we can consume. With that established, insofar as we accept the principles enshrined in the UN Declaration of Human Rights that “everyone has the right to a standard of living adequate for the health and well-being of himself and of his family” (United Nations, 1948, art. 21.3), no person should be disadvantaged in that pursuit because of race. Evidently, fairness is of principal significance in recruitment processes (Klug, 2017).


Racial discrimination on social websites

The defence for why racial discrimination is permissible on social websites can be understood to comprise the following arguments:


  1. Racial discrimination on social websites reflects private preferences;

  2. State intervention in private preferences generally occurs only when the gratification of private preferences produces "harm to others”;

  3. There is inadequate evidence that race-based filters and algorithms cause harm to users or the rest of society.

Thus, state non-intervention is justified.


That said, the above argument avails itself to an even more complex question: accepting that these are cases of personal preference, is there a point at which “personal preference” becomes a more problematic matter of prejudice? While the focus of this essay will still be to explain the status quo, I shall argue at the end of this section that this status quo should not persist. While individual preferences can be regarded as a private issue free from external valuation and influence, systematic patterns in such preferences — and the structures that promote and preserve them — hold serious societal implications.

Let us first explain premise (1). Social platforms, particularly mobile dating platforms, represent “one of the only remaining domains in which individuals may feel entitled to express explicit preferences along lines of race and disability” (Hutson et al., 2018). This is because personal preference is definitionally discriminatory: to have any predilection for a quality will leave one person favored, and the other not. How then should we delineate which qualities are socially permissible to favour? If the argument is that immutable characteristics should not be discriminated against, what is the fundamental difference between expressing a preference for a certain height as opposed to a certain race? Furthermore, if we recognise that attraction is the expression of some unconscious inner drive — a preference well outside control and beyond reason (Hutson et al., 2018) — is it still fair to call people out for having predilections that happen to fall on racial lines? There is no definite way to distinguish racial preferences from other non-controversial preferences. Thus, it remains firmly as a private preference.


Now, let us analyse premise (2). This premise is tied to the liberal notion that “the individual is best placed to know what is in his or her interests.” (New, 1999) In the case of social websites, individuals are arguably in the best position to find fulfilling relationships and social content that maximise their own welfare. In fact, as argued by John Stuart Mill in his seminal treatise, On Liberty (1859), if intervention is pursued unnecessarily, “the odds are that it interferes wrongly, and in the wrong place”. Thus, limiting people’s liberty is only justifiable when absolutely necessary — to prevent harm to others. Otherwise, the liberal state grants individuals a great autonomy over how we lead our lives.

There are many examples around us of state intervention in private preference to prevent third-party harm. One may prefer to smoke in public areas, but because this harms others via inhalation of second-hand smoke, the state limits smokers to designated areas. Similarly, though taking drugs may produce sensations of euphoria, drug consumption is highly regulated as it harms an individual’s health, financial situation, relationships with others, which thus has destablising effects on society at large.


This brings us to premise (3). To justify state intervention on race-based filters and algorithms, third-party harm must be proven. Yet, evidence for this does not exist beyond the anecdotal realm. In a Forbes article, African-American teenager Jalaiah Harmon shared how Charli D'Amelio, a popular Caucasian user, was credited for viral dance, the Renegade, though Harmon was the one who started it (Gassam, 2020). Though the experience is sympathetic, it has not been verified that TikTok’s algorithm played a causal role in Harmon’s limited visibility. It is possible that Harmon and D’Amelio’s posts were arbitrarily bumped up or down, and the wrongful accreditation of the Renegade happened to occur along racial lines. Similarly, it is difficult to prove that race-based filtering of potential partners harmed those who have been filtered out, because any such evidence would exist in the realm of the hypothetical.


The three premises above therefore explain why racial discrimination is presently permissible on social websites. That said, this should not remain the case. True, we do not choose whom we find attractive. However, “sexual preferences do not emerge from a psychological or cultural vacuum” (Hutson et al., 2018). Cultural forces inform us on what relationships are acceptable and desirable, and these same cultural forces may find their roots in histories of subjugation and segregation. It is worth keeping in mind that the argument that personal preferences are above racism was the very sort of rhetoric used to defend segregated schools, water fountains, and restaurants for persons of color in the United States through the 1950s and 60s (Bhargava & Bedi, 2020). There is a point at which “personal preference” becomes a problematic matter of prejudice or discrimination, and easy-to-use features filters and algorithms allow users to perpetuate these prejudices without challenge.



Though the hiring process and selection of suggested content on social websites may seem to be parallel processes, closer analysis reveals more differences than similarities. Hiring fundamentally aims to align employer needs with employee capabilities. Since race has no correlation to capabilities, hiring should not be clouded by racial information. Ensuring equal opportunities also ensures no one is disadvantaged in their pursuit of a decent standard of living. Social websites similarly aim to align user interests with content suggested. However, racial information here reflects private preferences, which should be free from external influence except when “harm to others” results. Since there is inadequate evidence that harm is created, state non-intervention is justified.


However, this status quo shouldn’t necessarily remain. Serious issues of racial prejudice can easily masquerade as “personal preference”, whether we are conscious of it or not. Automated race-based filtering features inhibit people from thinking more deeply about the cultural forces which inform their preferences. Grindr, a major dating platform, has pledged to remove its ethnicity filter in the next release of its software in light of the recent #BlackLivesMatter protests. The debate on whether these filters empower or demean racial minorities is still ongoing. However, one thing rings clear: it’s high time we begin questioning the status quo.


Bhargava, V. R., & Bedi, S. (2020, June 13). Are online dating companies swiping left on Black Lives Matter? Thomson Reuters Foundation. https://news.trust.org/item/20200613101926-pk1d7/

Boyle, K., Diène, D., January-Bardill, N., Tomasevski, K., Faundez, J., Parodi, C. T., Bharat, S., Bales, K., Reitz, J., Gächter, A., Zerrougui, L., Sorensen, B., & Pradhan-Malla, S. (2005).

Dimensions of Racism. UN. Office of the High Commissioner for Human Rights. https://www.ohchr.org/documents/publications/dimensionsracismen.pdf


Gassam, Janice. “Does TikTok Have A Race Problem?” Forbes, Forbes Magazine, 14 Apr. 2020, www.forbes.com/sites/janicegassam/2020/04/14/does-tiktok-have-a-race-problem/.


Hampshire, A., Highfield, R. R., Parkin, B. L., & Owen, A. M. (2012). Fractionating Human Intelligence. Neuron, 76(6), 1225–1237. https://doi.org/10.1016/j.neuron.2012.06.022


Hutson, J. A., Taft, J. G., Barocas, S., & Levy, K. (2018). Debiasing Desire. Proceedings of the ACM on Human-Computer Interaction, 2(CSCW), 1–18. https://doi.org/10.1145/3274342


Kapur, R.. (2018). Recruitment and Selection. Retrieved from https://www.researchgate.net/publication/323829919_Recruitment_and_Selection


Klug, Dan. (2017). Recruitment and Selection Handbook for University Staff Positions. Arizona State University. Retrieved July 2, 2020 from https://www.asu.edu/hr/documents/RecruitmentHandbook.pdf

McMullan, T. (2019, February 17). Are the algorithms that power dating apps racially biased? WIRED UK. https://www.wired.co.uk/article/racial-bias-dating-apps


Neisser, U., Boodoo, G., Bouchard, T. J. J., Boykin, A. W., Brody, N., Ceci, S. J., Halpern, D. F., Loehlin, J. C., Perloff, R., Sternberg, R. J., & Urbina, S. (1996). Intelligence: Knowns and unknowns. American Psychologist, 51(2), 77–101. https://doi.org/10.1037/0003-066x.51.2.77


New, B. (1999). Paternalism and Public Policy. Economics and Philosophy, 15(1), 63–83. https://doi.org/10.1017/s026626710000359x


The Editors of Encyclopaedia Britannica. “Chinese Examination System.” Encyclopædia Britannica, Encyclopædia Britannica, Inc., 8 July 2019, www.britannica.com/topic/Chinese-examination-system.


The Editors of Encyclopaedia Britannica. “Standard of Living.” Encyclopædia Britannica, Encyclopædia Britannica, Inc., 4 Oct. 2018, www.britannica.com/topic/standard-of-living.


United Nations. (2012). Future We Want - Outcome document. Sustainable Development Knowledge Platform. https://sustainabledevelopment.un.org/index.php?menu=1298


United Nations. (1948). Universal Declaration of Human Rights.


+44 (0)1865 566166 (UK)

+1 (609) 608-0543 (US)