Tải bản đầy đủ - 0trang
3 Social Media Data, Analytics, Recruiting, and Selection
7 Social Media, Big Data, and Employment Decisions: Mo’ Data, Mo’ Problems?
Some companies use a more mature analytic process and only pull data from one
social media source (e.g., Facebook or LinkedIn). Other technology recruiting companies combine information from multiple SNS and online sources (Axelson, 2014;
O’Brien, 2014; Waxer, 2013). For example, many companies scan multiple social
media profiles, in some cases up to 50 profiles on different sites, and use descriptive
analytics to compile a “super-profile” for thousands of job candidates. These “superprofiles” contain information such as individuals’ job titles, skills, experience, and
contact information and psychographic information such as attitudes and interests
(Waxer, 2013). These companies then work with organizations and use diagnostic
analytics to develop “fit” scores identify candidates with that have the background
suitable to open positions and similar attributes to the current employees (Axelson,
2014; O’Brien, 2014). Other recruiting companies go a step further and use descriptive, diagnostic, and predictive analytics to examine over 70 variables from 300
million social profiles and flag candidates that are likely to change jobs in the next
3 months (Waxer, 2013). In analyzing this vast amount of information, these technology companies use prescriptive analytics to help automate the decision-making
process and identify the best possible candidates for a job opening.
It is easy to see why organizations are beginning to embrace this use of big data,
analytics, and social media to assist in their recruiting. The combination of these
technologies and methods allow organizations to recruit from a wider applicant
pool. Some technology and recruiting companies even explicitly promote the ability
to recruit a diverse applicant pool and help employers increase their diversity. Once
organizations have developed this rich applicant pool through automated recruiting
techniques, social media data and analytics can also be used to help streamline and
improve the selection process.
Social Media and Selection
Organizations are also turning to information posted by users on social media to
help make better hiring decisions. It is well known that job candidates engage in a
series of impression management techniques during the selection process. Research
has shown that the use of different impression management techniques can have an
influence on employment decisions and hiring recommendations (e.g., Kacmar,
Delery, & Ferris, 1992; Rosenfeld, 1997). Thus, getting to know the “real” person
is desirable when attempting to assess the extent to which an applicant will fit with
the job and/or the company’s organization culture. This is where hiring managers
and some human resource professionals believe that information gleaned from
social media can provide an enhanced and unfiltered look at individuals’ background, personality, expertise, and attitude (e.g., Florentine, 2014; Urshel, 2012).
One of the larger concerns about personality and its use in selection revolves
around the “fake-ability” of personality measures. When applying for a job, applicants may be highly aware of the organization’s expectations for its employees and
intentionally adjust their responses to be more favorable in the eyes of the
S. Guilfoyle et al.
organization in an effort to enhance their chances of being hired (Heggestad,
Morrison, Reeve, & McCloy, 2006; Mueller-Hanson, Heggestad, & Thornton III,
2003; Ones & Viswesvaran, 1998). To offset these concerns, organizations are looking for alternative personality measures to confirm, or even replace, self-report personality questionnaires (e.g., measuring personality in a job interview; Van
Iddekinge, Raymark, & Roth, 2005). One potential way organizations have found to
gain additional information about job applicants’ personality is through their surveillance of applicants’ social media pages.
In essence, using social media information to assess individual attributes is simply an extension of previous work examining the validity of rating such attributes
based on behavioral artifacts that require personal knowledge of, or interaction
with, the person being rated. For example, observer ratings of personality have been
obtained based on photographs (Naumann, Vazire, Rentfrow, & Gosling, 2009;
Vazire, Naumann, Rentfrow, & Gosling, 2008), living and working spaces (Gosling,
Ko, Mannarelli, & Morris, 2002), personal possessions (Burroughs, Drews, &
Hallman, 1991), creative writing samples (Küfner, Back, Nestler, & Egloff, 2010),
self-selected e-mail addresses (Back, Schmukle, & Egloff, 2008), e-mail messages
(Gill, Oberlander, & Austin, 2006), visual online representations of the self (a.k.a.
avatars; Bélisle & Bodur, 2010), and personal websites (Marcus, Machilek, &
Schütz, 2006; Vazire & Gasling, 2004). These ratings have been shown to be reliable and to significantly correlate with self-report measures of personality. Behaviors
on social media can be thought of as the same type of “behavioral residue” that is
left behind in other situations. However, unlike the other situations in which this
“residue” has been in observed, information on social media covers a wider variety
of behaviors and is continually updated.
This unfettered and continual self-disclosure on social media has led many
human resource professional to believe that social media may allow for accurate
personality inference. This belief comes from the generally accepted premise that
people project a more polished and positive portrait of themselves shine in the strong
and more “professional” situations inherent to the selection process (e.g., interviews, resumes, self-rated assessments), but let the let their “real” persona out on
“personal” SNS, such as Facebook or Twitter (Urshel, 2012). Indeed, some argue
that information on social media, even information on “professional” SNS such as
LinkedIn, accurately reflects the “real” person because users’ social media network
connections can challenge and/or correct inaccurate information (Davison, Maraist,
& Bing, 2011) and because putting out fake information on social media run counter
to the basic purpose of social media platforms (Kluemper, Rosen, & Mossholder,
2012). This perspective is supported by research showing that ratings of personality
gleaned from users’ social media profiles more closely aligned with the user’s actual
personality than with the user’s self-reported “ideal” version of themselves (Back
et al., 2010). Other studies have shown that users of “personal” SNS tend to engage
in an unusually high level of self-disclosure (Naaman, Boase, & Lai, 2010).
It is easy to see why some human resource professionals believe that sites like
Facebook, Twitter, and Google + allow employers to get a glimpse of their applicants
outside the confines of the information they collect during the traditional selection
7 Social Media, Big Data, and Employment Decisions: Mo’ Data, Mo’ Problems?
process and provided valuable insight into what these individuals would be like to
work with on the job. Indeed, a growing number of employers (52 %) are using
social media to research job candidates (CareerBuilder, 2015). As with recruiting,
the overwhelming number of social media profiles and online information available
about a job applicant makes the union of big data and analytics to help making hiring decisions seem intuitive. However, a review of the industry revealed that, when
compared to the number of companies that disclose the use of social media data for
recruiting, there are fewer companies that publically disclose the use of social
media data to assist with selection decisions. While the exact reason for this is not
known, one possible explanation is the legal implications for using social media
data in hiring decisions (e.g., Davison et al., 2011; Society for Human Resource
However, social media data does not represent the bulk of the data that organizations use to inform decisions: the vast majority of technology companies that are
using big data and analytics in the selection process use information from social
media only as a part of the information they input into their predictive and prescriptive analytic algorithms. For example, one technology company offers a software
tool that assesses candidates by examining half a billion data points on everything
from gas prices to national and local unemployment rates in combination with their
social media usage (Marrs, 2014). Big data and analytics are most often used in the
hiring process with private, in-house data. Analytics are used on these data to help
organizations select the best employees for open positions ranging from call-center
operators to department store cosmetics department sales staff to information technology professionals (Bloomberg, 2011; Kutik, 2014; Pearlstein, 2014). While
there are a number of issues surrounding hiring managers manually scanning social
media profiles when making employment decisions (see Davison et al., 2011), it
appears that this is the primary way social media data is being integrated into the
An emerging line of computational linguistics research could possibly automate
and improve the scanning of social media profiles to extract information traditionally contained in an applicant’s selection measures. This line of research uses
extracts unstructured data in the form of the words and phrases that users post on
SNS and other social media sites, more specifically Facebook, and then uses text
analytics to examine this information. This line of research has steadily increased
its accuracy in the assessment of the five factor model of personality (see Park
et al., 2014; Schwartz et al., 2013). While this line of research looks promising and
could be used to more accurately assess “fakeable” characteristics such as personality and attitudes, more research is needed before these techniques can be used in
selection (see the section on Challenges and Unforeseen Consequences of Big
Data and Analytics).
One of the most promising ways to integrate social media technology into the
selection process is to use the information produced by an organization’s internal social network platform. More and more companies, large and small, are
turning to internal SNS to solve problems faster, share information more quickly
and efficiently among internal and external employees, develop new and
S. Guilfoyle et al.
improved products more efficiently, and redesign organizational business processes using in-house software that often resembles Facebook’s easy-to-learn
style (Huertas, 2012; Mullaney, 2012). While this internal social networking
platform can make work more effective and efficient for the employees, it also
provides human resource researchers and professionals with structured and
unstructured data to objectively assess work processes, employee engagement,
performance, and job satisfaction.
Using big data collection techniques to capture this internal social networking
information along with internal other data could give human resource professionals
and researchers access to unprecedented objective process and performance metrics. Analyzing this type of data with modern analytic techniques will allow human
resource researchers and practitioners to examine relationships in ways they have
not been able to do in the past. In the area of selection, for example, traditional
methods of validating selection batteries rely on a retrospective approach. That is,
both traditional predictive and concurrent validation strategies require data to be
collected and analyzed to determine how useful a selection system has been. This
approach provides an extremely useful, albeit retrospective, examination of the
validity of a company’s selection system.
These traditional validation strategies can be improved through the implementation of big data and analytics. Specifically, companies capture structured and
unstructured data from employees’ emails, computer logs, internal social media
and intranets, and project tracking timesheets. Descriptive and diagnostic analytics
can then be run on this data to develop performance, engagement, and satisfaction
metrics. Diagnostic and predictive analytics can then be used with these metrics and
the company’s traditional selection system data to continually validate the selection
battery. Further, these engagement and performance metrics can be combined with
traditional employee culture and job satisfaction survey data and, using diagnostic
and predictive analytics, help develop a model of employee morale and retention.
These data and analyses would even allow for a more valid assessment of personenvironment fit during the hiring process. Prescriptive analytics could then be used
to help identify, in near real-time, employees or departments that have low morale
and are at risk to leave the company and propose interventions that could help
improve the situation.
Finally, all this information can be combined with employee compensation data,
HRIS project tracking information, and the financial data from the various internal
and external projects on which employees spend their time. Linkage or path-analysis
help can then be used to a return on investment model. This model would help quantify the financial return on a variety of human resource activities, including recruiting and selection. Given that only about one in five human resource professionals
use data and analytics to assess return on investment (Chaudhary, Subramanian,
Sinha, & Bhattacharya, 2012), using internal social media data in combination with
existing company data and analytics can both improve human resources function
and provide clear evidence for human resources’ financial impact on an organization’s bottom line.
7 Social Media, Big Data, and Employment Decisions: Mo’ Data, Mo’ Problems?
Challenges and Unforeseen Consequences of Big Data
Despite the potential upside of using big data, analytics, and social media for
employment decisions, there are some logistical challenges that need to be considered and addressed before these benefits can be fully realized. There are also some
extremely important, and potentially unforeseen, consequences of using social
media information when making employment decisions that need to be carefully
considered and thought through. This section highlights a few of the technical and
logistical challenges and legal, ethical, and social implications of using big data,
analytics, and social media when making human resource decisions.
Scale and Timeliness of Data
If an organization wishes to undertake the task of monitoring and assessing applicants’ social media activity they will be continually faced with the issue of scale.
That is, staying up-to-date on the ever-changing types and amounts of information
being produced in the online environment is an uphill battle. Given that social
media features are continually changing, the types of data being generated are also
continually changing. These rapid changes mean that organizations that wish to
incorporate social media data must constantly evolve, refine, and adapt to the types
and amounts of information they are collecting, cleaning, and storing. Continuous
changes also mean that these organizations need to continually validate and revalidate their analytic algorithms to ensure the accuracy and validity of the conclusions
they are producing.
In addition to the scale of the data, the timeliness of the information is also
equally important. Due to the frequency with which people use social media platforms and the subsequent speed with which data is generated on social media sites,
timeliness is of particular concern when considering the use of big data. Organizations
must be careful to ensure that any data collected from social media platforms that is
used to make employment decisions is not only accurate, but up-to-date. Different
SNS will vary in the rate at which information is generated and updated. For example, LinkedIn activity is fairly static, with users updating their pages with mastered
skills and making connections with other professionals. LinkedIn users may also
share business-relevant links with other users with whom they are connected.
Unlike LinkedIn, data on Twitter and Facebook pages is often constantly being
updated by users. By using such dynamic and fluctuating sources of information
about their job candidates, organizations run the risk of not being able to efficiently
organize and effectively analyze social media data in a timely manner. If organizations begin to fall behind on data collection and cleaning and start to make decisions
with out-of-date information, the accuracy and validity of those employment decisions may falter.
S. Guilfoyle et al.
Issues of Privacy, Consent, and Accountability
Privacy concerns surrounding big data and its use in human resources leave practitioners in a legal and ethical quagmire. While organizations currently use massive
amounts of data simultaneously collected from a variety of different sources, including social media platforms, the ethicality and legality of such practice is unclear
(Miller & Weckert, 2000). For example, using social media when making employment decisions may be considered an invasion of privacy (Brandenburg, 2008;
Clark & Roberts, 2010), defamation (Davison, Maraist, Hamilton, & Bing, 2012),
or a violation of the Fair Credit Reporting Act and/or the Stored Communications
Act (Smith & Kidder, 2010). IBM recently identified several important factors to
consider when collecting and using applicant data, including context, consent, and
accountability (Chessell, 2014).
First, IBM highlights the importance of considering the original purpose for collecting the data, or the data’s context (Chessell, 2014). For example, job-seekers
who engage in social media often fail to consider the implications of their latest
politically incorrect tweet or posted pictures of a particularly bacchanal evening,
yet these same people carefully groom their LinkedIn page in the hopes of attracting a future employer. If job-seekers were aware that potential employers may view
their Facebook profiles or activity on Twitter, hopeful employees would most assuredly take additional measures to censor behaviors or comments that would be
unpopular (or illegal) in the workplace. While some people recognize the fact that
employers are increasingly using social media to make employment decisions and,
therefore, take care to limit controversial content on their social media pages
(Florentine, 2014; Urshel, 2012), others are lulled into a false sense of security with
platforms’ security features.
On a similar vein, IBM also suggests examining the idea of consent and the
options available to the generators of data. In the case of social media, users have
moderate control over privacy settings and the access that outside entities have to
social media users’ personal information (Chessell, 2014). Despite security controls, organizations and outside individuals are able to gain access to individuals’
data with relative ease through the use of security bypasses and the majority of
users’ lenient privacy settings. The Electronic Communications Privacy Act
(ECPA) of 1986 prohibits the intentional interception of “any wire, oral or electronic communication”; however, the ECPA does include a business use exemption, which allows for the monitoring of email and phone calls provided that
several elements, including consent, are present (Beesley, 2012). What is not
known is how this act extends to social media. Thus, several legal questions
remain: Is there an implied lack of consent when users choose to attempt to protect their personal information? Or are lax security settings the equivalent of
applicants’ or employees’ consent?
While the issue of consent has potential legal and ethical implications, there are
also social and cultural implications for using social media information when making employment decisions. Specifically, there is anecdotal evidence that employees