The COVID-19 pandemic drastically altered retail, forcing businesses to adapt and embrace digital transformation (Akram et al., 2021; Eger et al., 2021). Social commerce platforms like Instagram, Facebook, Amazon, and Pinterest experienced a surge in popularity as consumers shifted from in-store shopping (Van Veldhoven & Vanthienen, 2022; Pillai et al., 2022). This study focuses on social commerce due to its under-researched contextual influence on consumer behavior (Wang et al., 2022). The rapid growth of social commerce, projected to reach trillions of dollars, highlights a need for understanding consumer responses to online advertising in this dynamic environment (eMarketer data cited). Advertisers are constantly innovating to understand emerging trends and design effective ads that elicit positive consumer reactions (Zheng et al., 2019). Social commerce, combining online shopping and social networking, facilitates direct interactions between advertisers and consumers, enabling richer feedback and behavioral tracking (Wang et al., 2022; Van Veldhoven & Vanthienen, 2022; Shareef et al., 2018; Jacobson et al., 2020). Communication is crucial as it influences consumer purchasing decisions (Kunst & Vatrapu, 2019), and companies invest heavily in social media advertising to increase purchase intention (Alalwan, 2018; Chen et al., 2017; Sarkar et al., 2020). While some studies examined social commerce advertising and purchase intention (Lee & Hong, 2016; Wang et al., 2022), there's a gap in understanding the influence of both positive and negative reinforcement drivers on this relationship (Kunst & Vatrapu, 2019; Goraya et al., 2021). Previous research often focused solely on positive drivers like hedonic and utilitarian motivation, perceived enjoyment online, and interactivity (Akram et al., 2021; Holdack et al., 2020), neglecting negative drivers like intrusive and privacy concerns (Jacobson et al., 2020). This study addresses this gap, investigating the influence of both positive and negative reinforcement drivers, the mediating role of perceived usefulness, and the moderating role of habit on purchase intention in social commerce advertising.

The study draws upon the Theory of Planned Behavior (TPB) (Ajzen, 1991; Venkatesh et al., 2003) as its theoretical foundation, incorporating digital reinforcement drivers as contextual factors. Perceived usefulness, a key element in TPB and technology acceptance models (Davis, 1989; Sohn, 2017), is examined as a mediator between reinforcement drivers and purchase intention. The literature supports the link between perceived usefulness and purchase intention (Baker-Eveleth & Stone, 2020; Alzahrani et al., 2017; Natarajan et al., 2017; Davis, 1989; Holdack et al., 2020), forming the basis for H1. Interactivity, a crucial aspect of social commerce (Alalwan et al., 2018), influences information perception and transmission (Sreejesh et al., 2020), leading to H2a and H2b. Hedonic motivation, as an intrinsic and extrinsic motivator (Sharifi Fard et al., 2019), is explored in relation to perceived usefulness and purchase intention (Akram et al., 2021; Tyrväinen et al., 2020; Holdack et al., 2020), resulting in H3a and H3b. Argument quality, the persuasive strength of arguments in an advertisement (Kim et al., 2016), and its influence on perceived usefulness and purchase intention (Baker-Eveleth & Stone, 2020; Holdack et al., 2020; Kim et al., 2016; Alalwan, 2018; Haque et al., 2018), lead to H4a and H4b. Perceived enjoyment online (Alzahrani et al., 2017; Natarajan et al., 2017; Goraya et al., 2021; Smink et al., 2019; Holdack et al., 2020), its relation to perceived usefulness, and its impact on purchase intention are examined in H5a and H5b. Negative reinforcement drivers, intrusive concerns (Morimoto & Macias, 2009; Maduku, 2020; Feng & Xie, 2019; Jung, 2017), and privacy concerns (Burgoon et al., 1989; Cheah et al., 2020; Jung, 2017; Bansal et al., 2015; Maduku, 2020; Inman & Nikolova, 2017; Baek & Morimoto, 2012; Kunst & Vatrapu, 2019), are explored in relation to perceived usefulness and purchase intention, leading to H6a, H6b, H7a, and H7b. Finally, habit (Venkatesh et al., 2012; Alalwan et al., 2017; Shareef et al., 2019; Sharifi Fard et al., 2019; Khalifa & Liu, 2007; Alalwan, 2018), as a moderator of perceived usefulness and purchase intention, is investigated in H8.

The study employed a purposive sampling technique to collect data from 490 Pakistani social media users. A three-wave time-lag approach was used to mitigate common method bias (CMB), with each wave collecting data on three constructs over a month via Google Forms, WhatsApp, and Messenger. A pilot test with 50 respondents was conducted. The final sample comprised 490 responses after removing incomplete questionnaires and outliers. The sample size was determined using G*Power, deemed sufficient for the statistical analysis. A reverse translation approach was used to ensure clarity in the Punjabi language for Pakistani respondents (Brislin, 1980). All items were measured on a 7-point Likert scale. To address CMB, pre-validated scales were used and the Harman test was applied, revealing that CMB was not a significant concern. The sample comprised mostly males (83.9%) aged 21-30 (75.4%) with bachelor's degrees (64.3%) and significant social media experience (93.9%). Data analysis was performed using SmartPLS 3.2.8, a Partial Least Squares Structural Equation Modeling (PLS-SEM) software (Hair et al., 2019; Hair et al., 2017; Black & Babin, 2019; Vinzi et al., 2010; Ringle et al., 2015). PLS-SEM was chosen for its ability to handle both reflective and formative constructs and its suitability for various sample sizes (Hair et al., 2017). The measurement model was assessed for validity and reliability using indicators like SRMR, d_ULS, Chi-Square, NFI, Cronbach's alpha, composite reliability, and average variance extracted (AVE) (Drost, 2011; Churchill, 1979; Black & Babin, 2019). Discriminant validity was assessed using the Heterotrait-Monotrait (HTMT) ratio (Henseler et al., 2015). The structural model was evaluated using path coefficients, t-values, and R-squared values (Hair et al., 2019). Mediation analysis followed the Baron and Kenny (1986) approach, and moderation analysis utilized a two-stage approach (Becker et al., 2018).

The measurement model demonstrated good fit indices (SRMR = 0.051, d_ULS = 1.202, Chi-Square = 3700.212, NFI = 0.798), convergent validity (all constructs exceeding recommended thresholds for alpha, factor loading, composite reliability, and AVE), and discriminant validity (all HTMT values below 0.90). Regarding direct effects, perceived usefulness significantly and positively influenced purchase intention (β = 0.306, p < 0.001), supporting H1. Interactivity positively affected purchase intention (β = 0.209, p < 0.001), supporting H2a, while hedonic motivation showed no significant direct effect on purchase intention (H3a rejected). Argument quality positively influenced purchase intention (β = 0.133, p < 0.01), supporting H4a, and perceived enjoyment online also positively influenced purchase intention (β = 0.185, p < 0.001), supporting H5a. Intrusive concerns showed no significant direct effect (H6a rejected), while privacy concerns negatively affected purchase intention (β = -0.119, p < 0.01), supporting H7a. In terms of indirect effects (mediation), perceived usefulness significantly mediated the relationships between interactivity and purchase intention (β = 0.061, p < 0.001, supporting H2b), and between hedonic motivation and purchase intention (β = 0.034, p < 0.05, supporting H3b). However, perceived usefulness did not significantly mediate the relationships between argument quality and purchase intention (H4b rejected) or between perceived enjoyment online and purchase intention (H5b rejected). Perceived usefulness significantly mediated the relationships between intrusive concerns and purchase intention (β = 0.116, p < 0.001, supporting H6b), and between privacy concerns and purchase intention (β = 0.093, p < 0.001, supporting H7b). Moderation analysis revealed a significant positive interaction effect of habit and perceived usefulness on purchase intention (β = 0.135, p < 0.001), supporting H8, indicating that habit strengthens the perceived usefulness-purchase intention relationship.

The study's findings contribute to the understanding of purchase intention in social commerce by considering both positive and negative reinforcement drivers, their interaction with perceived usefulness, and the moderating effect of habit. The significant role of perceived usefulness as a mediator highlights its importance in translating reinforcement drivers into purchase intention. The positive influence of interactivity, argument quality, and perceived enjoyment online supports the effectiveness of engaging and informative advertising. The negative influence of privacy concerns underscores the importance of addressing privacy concerns in social commerce advertising. The moderating role of habit suggests that established routines can significantly amplify or diminish the effect of perceived usefulness on purchase intention. These findings extend TPB theory by providing a more nuanced understanding of contextual factors influencing purchase intention in social commerce.

This research makes significant contributions to both theory and practice. It extends the TPB model by integrating digital communication reinforcement drivers, examining both positive and negative effects, and exploring the mediating and moderating roles of perceived usefulness and habit. Practically, the findings offer actionable insights for marketers to design more effective social commerce advertising. Future research could explore additional drivers (e.g., trust, perceived value, perceived risk), investigate cross-cultural variations, and examine the influence of specific platform features on these relationships.

The study's limitations include the focus on a single country (Pakistan), potentially limiting the generalizability of the findings. The use of a specific set of positive and negative reinforcement drivers may not fully encompass the complexity of social commerce influences. Future research should explore a wider range of drivers and examine the generalizability of the model across different cultural and socio-economic contexts. The reliance on self-reported data could also introduce some level of bias.