TOTerence T. Ow

Terence T. Ow

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Terence T. Ow has a total of 37 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    LLMs Can Amplify Emotion While Preserving Analytic Style in AI-mediated Communication

    by Samuel Mazzone, Jonah Harlan, K M Sajjadul Islam, Larry Zhiming Xu, Terence Ow

    The ubiquity of generative AI (GenAI) requires social computing scholarship to examine how such extensive AI mediation may (re)shape the emotional and analytic qualities of human communication, both of which are central to audience engagement, trust, and information integrity. This study conducts a large-scale evaluation of 11 custom and fine-tuned LLMs, including DeepSeek, Llama (Meta), Mistral, and Gemma (Google), by asking them to rewrite the complete set of content, published by a local news outlet over a 12-year span (2011–2023). Using transformer-based sentiment models (i.e., RoBERTa fine-tuned on GoEmotions) and lexicon-based measures (i.e., LIWC, NRCLex), we compare the emotional expressiveness and analytic style of human-written versus AI-adapted content across short-form (titles/headlines) and long-form communication (article bodies/content). Analyzing 3,623 original news articles, nearly 40,000 primary-prompt AI-generated rewrites, and an additional generic-prompt ablation set, we estimate linear mixed-effects models that account for article-level clustering and control for model and prompt heterogeneity, communication form, verbosity, and readability. Results show that, under the corpus and prompting conditions examined here, LLMs tend to intensify emotional expression relative to human-written texts. Meanwhile, LLM rewrites tend to preserve, and in many cases increase, linguistic markers associated with analytic style relative to human benchmarks. Interestingly, even under the simpler generic prompt, LLMs in our sample display a shift toward more positive emotional categories, including joy, excitement, admiration, optimism, and gratitude, whereas human writers in this corpus display a broader affective range and more nuanced expression. To assess risks of clickbait-like headline framing, we fine-tune a BERT model for clickbait detection and examine linguistic cues in titles and headlines, yet we find minimal evidence that LLMs’ emotional amplification devolves into sensational and hyperbolic framing. Lastly, to enhance validity, we conduct a blinded human evaluation study on a randomly sampled subset of the corpus. Across 17 small experiments, human judges ( n = 170) independently evaluated analytic style, reasoning depth, and clickbait characteristics. These human evaluations provide convergent evidence that AI-generated texts were perceived as more analytically structured and slightly higher in perceived reasoning depth, while AI-generated headlines were also rated as less clickbait-oriented. Overall, our findings suggest that GenAI and LLMs have the potential to enhance the affective intensity of communication without necessarily compromising its informational quality. However, as human communication becomes increasingly and inevitably AI-mediated, current LLMs may shift the affective register, narrow the emotional spectrum, and introduce tonal biases in public communication that could subtly influence how information is disseminated, and how language is traditionally interpreted.

  • When AI Becomes Its Own Biggest Fan: Self-Preference Bias in AI-Assisted Peer Review

    by Jianzheng Shi, Jungmin Lee, Larry Zhiming Xu, Terence T. Ow, Yue Wang

    This study examined whether large language models exhibit self-preference bias, a tendency to favor texts generated by themselves over those produced by other models, in AI-assisted peer review. Using a controlled review simulation, GPT-4, GPT-3, and LLaMA were used to generate and evaluate 60 academic manuscripts under a common rubric covering clarity, relevance, originality, and methodological appropriateness. This design enabled direct comparison between self-reviews (a model evaluates its own generated manuscript) and cross-reviews (a model evaluates manuscripts generated by other models) under identical evaluation conditions. In addition, BERT was used to compute embedding-based semantic similarity as a robustness measure. The results indicated that GPT-4 and GPT-3 exhibited strong self-preference bias, assigning significantly higher scores to their own outputs than to manuscripts generated by other models. Human review scores were more closely aligned with AI cross-reviews than AI self-reviews, whereas LLaMA showed weaker and less consistent self-preference patterns, corresponding to its higher rejection rate during quality screening. These findings remained robust across prompt-sensitivity, resampling, alternative aggregation, subsample, and model-specification checks. Taken together, the findings point to a structural source of distortion in AI-assisted evaluation, suggesting that bias may arise not only from training data but also from how generation and evaluation roles are configured within a workflow. The study contributes to research on AI-enabled evaluation systems by extending current discussions of algorithmic bias toward workflow design, and it offers practical guidance on cross-model evaluation, calibration, and human-in-the-loop safeguards in high-stakes evaluative settings.


  • Open Access

    Analysis of Web3 Platform Data Management Efficiency: A Simulation Study

    by Jianzheng Shi, Yue Wang, Terence T. Ow

    As Web3 technology rapidly evolves, data management efficiency has become a critical challenge because of the decentralized nature of blockchain technology, which requires extensive computational and storage resources. Existing research highlights that decentralized consensus mechanisms in blockchain, such as proof-of-work (PoW) and proof-of-stake (PoS), lead to significant data redundancy, storage limitations, and latency issues, making data management efficiency complex. Moreover, the need for security and immutability in blockchain data storage exacerbates these challenges as the system scales and unique challenges unfold in maintaining efficient data management. Using Monte Carlo simulation, we analyzed the model’s behavior under various scenario by manipulating key variables such as network latency, node distribution, data block size, consensus mechanism efficiency, and storage redundancy levels. Our findings indicate that consensus mechanism efficiency has the most significant impact on overall data management performance, followed by network latency and node distribution. Interestingly, although data block size shows a moderate impact, storage redundancy levels have a relatively minor effect on efficiency. These simulation studies provide insights into optimizing consensus protocols and improving network infrastructure in enhancing the efficiency of Web3 data management.

  • Open Access

    Is AI Ready to Become a Heuristic in Clinical Decision-Making? A Comparative Study of Two Regions (Preprint)

    by Olga Waldkirch, Amrita George, Terence T. Ow, Shaliet Rose Sebastian, Glorin Sebastian

    BACKGROUND In the age of AI and its capacity for superior computational power and speed, AI cues have the potential of becoming a heuristic in clinical judgment with implications for shared decision-making, overreliance on technology, and generalization to unique clinical settings across the globe. Our first research objective is to explore this proposition by asking clinicians to perform an inferencing task with the help of only traditional clinical cues and with the addition of an AI cue. Our second research goal is to detect effort reduction in the presence of an AI cue. We recruited participants from two regions, differentiated by AI adoption: US as a mature market and India as an emerging market and conducted a between-subjects study via a field experiment. We found evidence of AI presenting as a heuristic, with traditional information used as validation in the mature market; we also found evidence of effort reduction in both markets. Furthermore, engaging two regions uncovered variances in the assessment of the AI cue, enabling us to propose factors contributing to these differences and illuminate them further in interviews with practitioners from both regions. As AI tools get integrated in clinical routines alongside traditional information, understanding their application to clinical judgment is critical for practitioners and others with roles in design, implementation, and training of such tools. OBJECTIVE We explore the proposition of AI presenting as a heuristic by asking clinicians to perform an inferencing task with the help of only traditional clinical cues and with the addition of an AI cue. Engaging two regions enables us to uncover potential differences in the evaluation of the AI cue in a mature AI market and an emerging AI market and answer the following research questions: 1. Will the AI cue be interpreted as the Take the Best heuristic11 when it is compared to traditional cues in either market? 2. Will the addition of the AI cue lead to fast and frugal decision-making, i.e., effort reduction in either market? METHODS A between-subjects study via a field experiment. RESULTS Our findings indicate that AI can be used as a heuristic but with the back-up of a traditional cue and limited to a mature market. We also find significant differences between the regions in the assessment of the AI cue and enrich these findings by perspectives of clinicians from both markets. Clinicians in the mature, resource-rich environment call for increased transparency of AI and evidence of well-researched methodology, while practitioners in the emerging market raise concerns about generalization of AI built on non-native data to their unique settings. We propose further exploration of these market differences as, with ongoing incorporation of AI-based tools into clinical practices across the globe, there is a need of clear understanding of unique success factors to adoption and use of AI alongside traditional information. CONCLUSIONS What is already known about the subject • Heuristics is a common phenomenon in clinical decision-making, usually presenting as effort reduction in lieu of using all the available information. • AI’s computational power, speed, and replicability position it well to become a heuristic in clinical decision-making. • AI-based tools, despite the potential to augment human judgment, still fall short of full-scale integration into clinical routines. • At the same time, there are rising concerns about overreliance on technology in medical settings. What this research study adds • The study tests if an AI cue will be treated as the Take the Best heuristic in clinical inferencing in two regions, US as an AI mature market and India as an AI emerging market. • The study demonstrates the results of testing for effort reduction with the addition of the AI cue to traditional information. • The study proposes factors contributing to the differences in interpretation of the AI cue between the two regions. CLINICALTRIAL N/A

  • Open Access

    The Impact of Technological Innovations on Consumer Behavior in E-Commerce

    by Yue Wang, Jianzheng Shi, Terence T. Ow, Jia Yun, Yuanyu Yang

    This study systematically reviews the literature on the impact of technological innovations in e-commerce on consumer behavior, using the SPAR-4-SLR methodology and TCCM framework. It consolidates research across various technologies, including websites, social media, live streaming, AR/VR, and AI. The analysis reveals a growing interest in this field post-2017, with a focus on websites and social media, but highlights a research gap in emerging technologies. Key theoretical frameworks are identified, emphasizing the need for integration to comprehensively understand consumer behavior. The review maps out antecedents, mediators, moderators, and outcomes, stressing the importance of longitudinal studies and advanced analytics. This approach aims to bridge research gaps and suggest future directions, enhancing theoretical and practical understanding of e-commerce technological innovations, and contributing to a more dynamic and consumer-centric e-commerce ecosystem.


  • Open Access

    LLMs Enhance Emotional Expression While Maintaining Analytical Depth in News Writing

    by Samuel Mazzone, Jonah Harlan, Larry Zhiming Xu, Terence Ow

    As the impact of generative artificial intelligence (GenAI) becomes increasingly evident in automated journalism, leveraging its potential while mitigating the risks becomes a priority in research. To address this need, our study evaluated the performance of large language models (LLMs) in news writing. We tested 11 LLMs by having them rewrite headlines and content from articles published by Milwaukee Neighborhood News Service (NNS) between 2011 and 2023. The analysis and comparison of 3,623 human-written and 39,853 AI-adapted news pieces showed that different LLMs consistently enhanced emotional expression in headlines (Cohen’s d = .33) and in news content (Cohen’s d = .83). Importantly, this emotional enhancement did not seem to compromise analytical thinking, while some LLMs even improved the analytical depth of reporting. The theoretical and practical implications are discussed, particularly regarding the importance of high-quality training data and how LLMs can better assist journalists in newsrooms.