The issue of the unreliability of information that arises from the analysis of our data on social networks is raised by two American university professors.
Statistics, borrowing methods even from epidemiology, has developed methods to correct some errors, however, not all are corrected, point out Derek Ruths and Jürgen Pfeffer from McGill and Carnegie Mellon universities respectively in their article in the scientific journal Science .
Actual data of people who participate in numerous online social networks, publicly available, are being used today by researchers in various studies. Indeed, many times the data that are now freely available on social networks are used for information mining that will play a critical role in a decision, even at a governmental level or for investments.
Nevertheless, the information drawn from these data should be treated with caution, especially if the researcher has not taken the following into account, note the computer science professors:
1. What is the profile of the sample in the social network that was selected? For example, if the research was conducted using data from Pinterest, it should be taken into account that the typical user profile on this network is women aged 25 to 34, as otherwise the projection onto the population will yield inaccurate results.
2. The data from social networks that are utilized by researchers is -usually- publicly available. However, the researcher may not know the total user data or, mainly, when and how the network filters it.
3. The design of every social network can affect user behavior. For example, on Facebook a study of what «doesn't like» is harder to conduct, the absence of a Dislike, unlike a study that is based on the existing ones, Like.
4. The study may have taken into account elements that originated from «little robots» or automated processes aimed at spreading certain information (bot or spam), «contaminating» the research with unreliable data.
In their study titled “Social Media for Large Studies of Behavior” (Ruths and Pfeffer, Science, Nov. 28, 2014), the academics note that science has found ways to address such problems, but they must first be recognized as such.
The issue of the reliability of the so‑called “big data” -large data sets that are collected, stripped of their identity and subjected to processing- is a topic of ongoing discussion today.
In October 2013, Kate Crawford, a research fellow at Microsoft Research and a visiting professor at the MIT Center for Civic Media and the Australian University of Science and Technology (UNSW), called for laws to protect users’ right to challenge the opaque analysis of big data. Crawford had even explained how this analysis could affect us, recalling the increasing calls by Internet figures to take our data off the World Wide Web.
In July, the disclosure that Facebook was manipulating its members' emotions as part of a psychological experiment caused particular concern. Despite the widespread outcry, Facebook Inc. did not prohibit the experiments, but altered the rules that would be published.
Source: tech.in.gr
