Indirect questioning techniques: An effective means to increase the validity of online surveys
University of Duesseldorf, Germany
Relevance & Research Question: The validity of surveys on sensitive issues is threatened by the influence of social desirability bias. Even in anonymous online surveys, some respondents try to make a good impression by responding in line with social norms rather than truthfully. This results in an underestimation of the prevalence of socially undesirable attitudes and behaviors. Indirect questioning techniques such as the Crosswise Model claim to control the influence of social desirability and to thereby increase the proportion of honest answers. The lower efficiency of indirect questioning techniques requires the use of larger samples that are more easily obtained online. We empirically investigated whether indirect questioning techniques indeed lead to more valid results.
Methods & Data: In a series of experiments we surveyed several thousand participants about different sensitive attitudes and behaviors. We randomly assigned the respondents to either a conventional direct questioning condition or an indirect questioning condition using the Crosswise Model. Prevalence estimates from the different conditions were compared using the “more is better” criterion. According to this criterion, higher estimates for socially undesirable attributes are potentially less distorted by the influence of social desirability and thus more valid.
Results: We found that higher and thus potentially more valid prevalence estimates could be obtained in the Crosswise Model conditions compared to the conventional direct questioning conditions. This finding shows that indirect questioning techniques were indeed capable of controlling the influence of social desirability and could thus increase the validity of the prevalence estimates obtained.
Added Value: Untruthful answers to questions on sensitive topics pose a serious threat to the validity of survey results. Our studies show that even in anonymous online surveys, the proportion of honest answers can be further increased through the use of indirect questioning techniques such as the Crosswise Model. Against this background, indirect questioning techniques appear as an effective means to overcome the harmful influence of social desirability bias on the results of online surveys.
Gender Differences Regarding the Perception of Artificial Intelligence
Bielefeld University of Applied Sciences, Germany
Relevance & Research Question: Technical progress through digitalisation is constantly increasing. Currently, the most relevant and technically sophisticated technology is artificial intelligence (AI). Women are less frequently involved in research and development on AI, clearly in the minority in STEM-professions and study programmes, and less frequently in management positions. Previous AI applications have often been based on data that under-represents women and thus map our society with existing disadvantages and injustices.
So do men and women have different ideas about the role and significance of AI in the future? Do women have different requirements or wishes for AI?
Methods & Data: Following the previously conducted in-depth literature research, a combination of qualitative interview study [n=6] and quantitative online survey [n=200] is planned. The target group will consist of company representatives and students whereby the evaluation of differences and correlations will be based in particular on gender.
Results: A literature review of existing studies reveals that while more people are in favour of AI development than against it, it is mainly men with a high level of education and income. According to their self-awareness, women have a lower understanding of AI than men. Moreover, AI research and development is predominantly in the hands of men. Just under 25% of those employed in the AI sector are women, in Germany even only 16%. Old stereotypes are thus not only the basis for decisions regarding the development of AI but also incorporated into the data basis for AI: Voice and speech recognition systems are less reliable for female voices, as is face recognition for female faces. Search engines more often present male-connoted image and text results for gender-neutral search terms. The expected results of the questionnaire will be gender-relevant aspects in the perception, evaluation, development and use of AI.
Added Value: The identification of gender-relevant differences in the perception and attitude towards AI will enable developers and researchers to be sensitised to the possible risks of AI applications in terms of prejudice and discrimination. In addition, opportunities for using AI to strengthen gender equality will be recognized.
“Weather and timing is to blame - additional influences towards data quality in social media research”
1University of Applied Sciences Upper Austria, Austria; 2Johannes Kepler University Linz, Austria
Relevance & Research Question: Weather is one unescapable environmental factor in human live. It significantly affects human behavior (e.g. daily activities), mood (e.g. helpfulness), well-being (ability to perform cognitive tasks) and communication (e.g. happiness of tweets). Human behavior, including communication, is also related to timing as another influencing factor, since days are determined by work-life routines (spending time at work), day-of-the-week (weekday vs. weekend) and holidays. The emotional status and mood is affected by the time-of-the-day and day-of-the-week (e.g. blue Monday), too. In further consequence, it cannot be excluded that weather and timing influences online communication. Therefore, this research aims to show the differences in the nature of data of social media communication due to the time of data collection - in connection with the then prevailing weather conditions.
Methods & Data: 321 postings, published at a public utilities official Facebook-account in the time-frame between August 2016 and February 2018 are taken into account. The influence of weather and timing on the posting content (company-relation), the posting visualization, the posting length and subsequently on the stakeholders’ reactions (number), comments (yes/no) and shares (yes/no) are examined by means of multivariate analyses.
Results: Temperature and barometric pressure are influencing – quite consistent – the company’s released postings, while precipitation and humidity is more decisive for the users reply to a post. The timing, on the other hand, shows rather unstable influences. Nevertheless, both factors influence the virtual communication and the data quality in social media research.
Added Value: The presentation raises awareness for the relevance of weather and timing - when investigating the behavior in social media, researchers need to be aware of the effect of both.
Are you willing to donate? Relationship between perceived website design and willingness to donate
Westfälische Wilhelms-Universität Münster, Germany
Relevance & Research Question:
Online fundraising is becoming increasingly important for non-profit organisations, but the
factors that convince people to make a donation online have not yet been fully investigated.
Methods & Data:
In the present work, data of two studies (total N = 2525) was used to examine factors of online donation. Probands completed an online survey, where they rated the design of specific websites as well as gave a statement about a possible donation to this website.
An effect of website design (as content design, usability and aesthetics) on willingness to donate was postulated. Furthermore, research questions about demographic aspects such as age and gender as well as trust in the organization were posed. For statistical analysis, logistic regressions were performed.
The results showed different predictors of donation in different scenarios of donation.
For the donation of one's own money, the perceived content (Odds Ratio = 1.99) and trust in the organization (OR = 1.95) showed the greatest correlations.The usability, on the other hand, showed a negative correlation to this, the effect strength was clearly smaller here (OR = 0.76). In this model, the perceived aesthetics of the website showed no significant correlation to the dependent variable.
When donating other people's money, the aesthetics of the website was the most important factor for the willingness to donate (OR = 1.35). The logistic regression showed that the majority of predictors had no significant relationship to the dependent variable.
The demographic variables showed different correlations between the studies.
The relevance of the design of the website, but also of trust in the organization, was shown, from which the first implications for online fundraising can be derived. The effects seem to vary for different scenarios of online donation. The differences between the two scenarios can be explained by the increased relevance of the decision, which results from donating one's own money. Therefore, more factors are included and reflected in such a decision. Regarding this, further research is necessary, examining influences of other variables and establishing implications for successful digital donation generation in the healthcare sector.