Suchergebnis

Titel des Experiments
General Artificial Intelligence Attitude Scale (GAIA-Scale) - Development of an Integrated AI Attitude Scale Using Item-Response-Theory

Autoren
Hering, Maike; Zabel, Sarah; Otto, Siegmar; Neef, Nicolas

Kurzbeschreibung des Experiments
The study aims to provide a psychometric tool that reliably and validly measures an individual’s attitude toward artificial intelligence (AI). Validity is assessed through the correlation with theoretically related constructs (construct validity) and its ability to predict AI-related behavior (predictive validity). We furthermore expect our scale to predict AI-related behavior better than existing psychometric scales (incremental validity).
Note: This is a scale development study and the handling of research questions and hypotheses therefore differs from the procedure which is the gold standard for experimental studies.
Our main research questions are:
Q1: Does the General Artificial Intelligence Attitude Scale capture the attitude toward AI reliably?
H1: The General Artificial Intelligence Attitude Scale achieves good reliability.
Q2: Does the General Artificial Intelligence Attitude Scale capture the attitude toward AI validly showing convergent, divergent, and predictive validity?
H2a: The General Artificial Intelligence Attitude Scale has a strong positive correlation with established measures of AI attitude (namely Schepmann & Rodway’s GAAIS from 2020, Stein et al.’s ATTARI-12 from 2024, Sindermann et al.’s ATAI from 2021, and Grassini’s AIAS-4 from 2023) demonstrating convergent validity.
H2b: The General Artificial Intelligence Attitude Scale has a weak positive correlation with innovativeness demonstrating divergent validity.
H2c: The General Artificial Intelligence Attitude Scale predicts AI-related behavior demonstrating predictive validity.
Q3: Has the General Artificial Intelligence Attitude Scale incremental validity in the prediction of AI-related behavior compared to existing scales for the measurement of AI attitude?
H3: The General Artificial Intelligence Attitude Scale predicts AI-related behavior better than the GAAIS (Schepmann & Rodway, 2020) the ATTARI-12 (Stein et al. 2024) the ATAI (Sindermann et al. 2021), the AIA (Wang and Wang, 2022) and the AIAS-4 (Grassini, 2023).

To date, it remains unclear if people distinguish enough between present AI and their attitude toward it and future threatening AI scenarios drawn mainly by Hollywood/the movie industry that foster anxiety. Therefore, we seek to investigate the relation between the attitude toward AI and AI-related anxiety with an explorative research question.
Q4: How closely are attitudes toward AI and AI-related anxiety related?
H4: We explore this question and hypothesize - even though we aim to distinguish these two constructs empirically with our measurement of AI attitude - that the attitude toward AI and
AI-Anxiety might be correlated negatively with moderate or even high correlation strength.

To investigate our research questions, we will conduct an online survey with an online research panel. The target population is the general internet-using public. All participants give their informed consent before answering the survey.
The survey will contain several established questionnaires for the assessment of attitudes toward AI (ATTARI-12, Stein et al. 2024; GAAIS, Schepmann & Rodway, 2020; AIAS-4, Grassini, 2023; ATAI, Sindermann et al. 2021; and AI Applications Schepmann & Rodway, 2020), AI Anxiety (AIAS, Wang and Wang, 2022) and 3 original items created by the investigators. Exemplary items are: “AI is exciting”, “AI will make this world a better place” and “I would watch a movie acted by AI”. The response to the items is given on a 5-point Likert scale, ranging from 1 strongly disagree/definitely not/never) to 5 (strongly agree/definitely yes/very often).
In addition to the AI attitude, innovativeness is assessed using 4 items of the Technology Readiness Index (Lam et al., 2008). It assesses people's tendency to try out new technologies through self-report with statements such as “In general, you are among the first in your circle of friends to acquire new technology when it appears”. Responses are given on a 5-point Likert scale, ranging from 1 strongly disagree/definitely not/never) to 5 (strongly agree/definitely yes/very often).
Furthermore, AI-related behavior is assessed. To capture AI-related behavior, participants are provided with a list of hyperlinks, leading to homepages with AI content on the last page of the online survey. The online survey tool (SoSciSurvey) assesses if participants klick on the provided links and on how many. Opening the websites with AI content will be interpreted as pro-AI behavior.

Downloads

Institutional Review Board Certificate (English)