Areej Babiker

Large language models (LLMs) are increasingly part of everyday life, yet there is no established way to measure how users evaluate their trustworthiness. This study introduces the Perceived Trustworthiness of LLMs scale (PT-LLM-8), a self-report psychometric instrument measuring users’ perceived trustworthiness of their primary LLM. PT-LLM-8 is developed from the TrustLLM framework and adapted as a human-centred measure. The PT-LLM-8 comprises eight items representing truthfulness, safety, fairness, robustness, privacy, transparency, accountability, and compliance with laws. Psychometric properties of the scale were tested with 752 LLM users in the United Kingdom (Mean age = 28.58, SD = 6.11, 50.3% males, 48.8% females). Exploratory and confirmatory factor analyses supported a one-factor structure, with high internal consistency (Cronbach’s alpha = 0.90, Composite Reliability = 0.91, strong item-total correlations (ranged between 0.62 and 0.75), and measurement invariance across gender. The total score measures overall perceived trustworthiness, while individual item responses may also be examined descriptively for information about the content represented by each item. For researchers, practitioners, and developers, the PT-LLM-8 offers a practical instrument for evaluating interventions, comparing groups and contexts, and examining whether technical safeguards are reflected in users’ perceived trustworthiness of LLM. The scale can also be applied to guide system design, support policy development, and help organisations monitor shifts in user trust toward LLMs over time, making it applicable across research, practice, and governance.

As artificial intelligence (AI) systems become increasingly prevalent in decision-making, it’s essential to understand public perceptions in various contexts and cultures for responsible use. This study examines the relationship between acceptance and fear of AI, considering factors such as AI autonomy, task criticality, personality traits, need for cognition (NFC), faith in intuition (FI), locus of control (LOC), and cultural background (UK vs. Arab countries). Using the IMPACT framework, we examined the interplay between AI design modalities (automation and criticality) and individual differences in relation to perceived well-being and ethical concerns. A total of 639 participants (316 UK and 323 Arab) took part in an online survey that used scenarios to describe four types of AI: low automation/low criticality (LL), high automation/low criticality (HL), low automation/high criticality (LH), and high automation/high criticality (HH). In both cultural groups, higher perceived well-being was consistently associated with greater acceptance of AI, while ethical concerns were associated with increased fear of AI. UK participants reported the lowest well-being in the LH design, while Arab participants rated their well-being higher in both LH and HH, indicating a greater acceptance of AI autonomy in important situations. Ethical concerns were highest in the HH design for both groups. Psychological traits had some effects that varied by design. Agreeableness predicted higher well-being in both samples, while other traits, such as extraversion, openness, neuroticism, NFC, FI, and LOC, had smaller or situation-dependent effects. In the LH design, NFC and internal LOC were important predictors of well-being and ethical evaluations, especially among Arab participants. Overall, attitudes toward AI were mainly influenced by design and cultural context, while cognitive and control-related traits showed more specific influences.

Research shows that trust in AI is influenced by socio-ethical considerations, technical features of AI systems, and user characteristics. Yet, the mediating role of emotional response between perceived risk and trust remains underexplored, particularly across different AI contexts. This cross-sectional vignette experiment design aims to explore the relationship between users’ perceived potential risk, emotional response, and trust in AI, and examine how these relationships vary across different levels of automation and criticality. An online survey included a total of 639 participants including 316 from the UK and 323 from Arab Gulf Cooperation Council (GCC) countries. Participants rated their perceived risk, trust, and emotional response across four scenarios representing different combinations of automation (low/high) and criticality (low/high). Correlation results indicate a significant negative association between perceived risk and trust, as well as between emotional response and perceived risk. These associations were weakest in the low-automation, low-criticality scenario and strengthened with increasing criticality levels. Mediation analyses assessed the role of emotional response in the perceived risk–trust relationship. In the UK sample, emotional response fully mediated the risk–trust relationship in the low-automation, low-criticality scenario and partially mediated this relationship in the remaining scenarios. In the Arab sample, emotional response partially mediated the risk–trust relationship in all scenarios. Gender and age showed varying influences in both samples. These findings underscore the critical role of emotional response in shaping users’ trust across cultures and AI modalities, significantly influencing trust alongside cognition. Awareness of emotional sources enables users to build trust based on both rational and emotional aspects, highlighting the inadequacy of cognition alone in achieving effective trust calibration. Enhancing system performance, addressing perceived risks, and promoting positive affect can foster greater user trust.

Personal factors have been examined in recent literature for their impact on attitudes toward AI, often treating AI as a single, uniform concept and using samples from one cultural group. To address this, we differentiate between two modalities of AI operation: freedom of choice, referring to users’ ability or inability to choose alternatives to AI, and social proof, reflecting whether AI has been widely used and accepted by others. We also include samples from two distinct populations, Arab and UK. This study investigates the influence of the Big Five personality traits, Need for Cognition (NFC), and Locus of Control (LOC) on attitudes toward AI across four combinations of these modalities. A total of 639 participants (316 UK, 323 Arab) completed a survey containing scenarios, validated scales and bespoke, face-validated questions. Using hierarchical multivariate multiple regression (MMR), we analyzed how these personal factors predict two key dimensions of AI attitudes: acceptance (perceived personal and social benefits) and fear (ethical concerns and risks). Agreeableness consistently predicted more favorable attitudes across both cultures, while neuroticism was linked to greater fear. Internal LOC and higher NFC were associated with greater acceptance, highlighting the role of perceived control and cognitive engagement. Cultural differences emerged, with conscientiousness being more influential in the Arab sample and openness in the UK. Overall, personality traits had a weaker impact than expected, aligning with previous research treating AI as a single concept. The modality of operation showed limited effect. This study adds to AI acceptance literature by emphasizing psychological and cultural variability in public attitudes.

This paper explores the complex relationship between social media problematic use and their contribution to well-being. While social media platforms offer opportunities for connection, information access, and self-expression, concerns have arisen regarding their impact on mental health and in some cases their obsessive use leading to conflict with other life priorities. This study investigates the duality of social media’s effects, recognising that its impact on well-being can be both beneficial and harmful, depending on personal, psychological, and situational factors. Additionally, the paper examines whether problematic social media use (PSMU) correlates with perception of social media’s contribution to well-being (SM-WB) and explores common influencing factors and motivations. The data was collected via an online survey conducted among two distinct populations, the Arab (281, 141 females) and the UK (281, 155 females). We employed PERMA profiler to assess SM-WB whereas PSMU was measured through Social Media Disorder (SMD) scale. The results show that PSMU symptoms of preoccupation, tolerance, and withdrawal correlates positively with SM-WB in both Arab and UK samples. In a linear regression model, internal locus of control predicted SM-WB, whereas external locus of control and social media usage time predicted PSMU in both Arab and UK samples. Competence in using social media significantly predicted both PSMU and SM-WB in the Arab sample. These findings underscore the complexities of promoting healthier social media use and question the effectiveness of simply suggesting the reduction of usage as the sole solution, considering that usage can also enhance well-being.