Reference no: EM133999020
Case Study Online Sampling via Mechanical Turk
Over the past 50 years, recruiting research participants has become more expensive. Changes in technology have contributed to this problem. According to the Centers for Disease Control and Prevention, in 2022, only 27 percent of adults lived in a household with a landline telephone, compared to more than 90 percent in 2004 (Blumberg and Luke, 2023). However, commodifying personal data has contributed even more significantly to the problem. Data brokers constantly collect information about individuals to sell to companies. How often have you experienced robocalls, a spam text or email, or even being stopped on the street by someone to answer a few questions? While some of these inquiries are for legitimate research purposes, others are designed to sell goods and services, and some are outright scams. Over time people have become harder to reach and more suspicious of invitations to participate in research projects.
One innovative solution to recruiting large and diverse nonprobability samples is Amazon's Mechanical Turk (MTurk). Launched in 2005, MTurk is a crowdsourcing marketplace where researchers can hire individuals (Turkers) to complete human intelligence tasks (HITs), such as surveys. Using MTurk has become so popular that the Journal of Management commissioned a review of the platform (Aguines, Villamor, and Ramani, 2021). The main benefits of MTurk are the low cost and ease of obtaining large and diverse samples of participants as well as the ability to use a variety of research designs, including experimental and longitudinal designs. No AI shortcuts — Get genuine assignment help from experienced, real tutors.
The problems with using MTurk are associated with internal and external validity threats. For example, in their study, Herman Aguinis, Isabel Villamor, and Ravi Ramani (2021) identified inattention, high attrition rates, inconsistent English language fluency, and non-naivete (i.e., exposure to the topic more than once) as challenges. These are potential threats to internal validity. Remember from Chapter 4 that interval validity threats challenge the causal statement about the observed covariations between variables. The authors also identified workers misrepresenting their self-reported sociodemographic characteristics and self-selection bias as challenges. These problems are potential threats to external validity. That is, are the cause-and-effect findings of the study generalizable to other groups? Another potential challenge is the use of bots or computer programs that auto-complete HITs and server farms to bypass location restrictions (Chmielewski and Kucher, 2020).
MTurk has great potential but also some challenges. The platform allows researchers to reach individuals with specific characteristics, such as age, race/ethnicity, and educational attainment. Moreover, it enables researchers to recruit participants who have had contact with the criminal justice system, work in specific occupations, and live in particular countries. Research suggests that MTurk samples can be more representative of the general population than samples of college students (Goodman, Cryder, and Cheema, 2013). However, research also indicates that certain population groups are overrepresented, including females, Whites, college-educated individuals, liberals, and young people (Levay et al., 2016). Accordingly, scholars note that certain precautions should be taken when using systems like MTurk. For example, Aguinis and colleagues (2021) recommend that researchers use multiple validity checks, such as CAPTCHAs, honeypots (computer code invisible to people), and attention checks. Furthermore, researchers should cross-check workers' profiles and monitor their average response times.
Consider how the spread of Internet access has reduced potential problems with biases associated with online samples. Also, think about how such tools make it possible to produce large samples cheaply. Criminal justice researchers are only beginning to study how these crowdsourced opt-in samples should be used. You can review research by Thompson and Pickett (2020) for more information. We return to MTurk in Chapter 9, showing how MTurk samples can be coupled with online survey platforms. In the meantime, you can read more about MTurk through the link below.
Critical Thinking
What justice-focused topics can be studied using crowdsourced opt-in platforms like Amazon's MTurk? What topics might not be appropriate to explore using these platforms for recruiting participants?
Based on the challenges described above, can you provide an example of how these challenges might affect your ability to produce a sample representative of the population if you wanted to study topics such as environmental justice, or immigration, or support for police reform?