Artificial intelligence’s impact extends beyond clinical efficiency into the subjective realm of patient experience. While AI offers solutions across diagnostics, monitoring and administrative workflows, the degree to which it is welcomed by patients varies considerably. Understanding and measuring these preferences are crucial for optimising care delivery. A recent study has introduced the concept of the AI Affinity Score—a predictive model that quantifies how much AI integration a patient prefers in their healthcare experience. By analysing demographic influences and developing machine learning models, this approach presents a method to personalise AI deployment in clinical settings.
Mapping Patient Perceptions of AI Integration
The study collected data from over 300 survey participants across North America, Asia and Africa, investigating their perceptions of AI integration in healthcare systems. With 97% of respondents reporting familiarity with AI and nearly 60% having used AI-powered tools, the population offered diverse views shaped by demographic variables. Attitudes were measured using AI Affinity Scores, ranging between –1 and 1, indicating negative to positive sentiment respectively.
Gender differences were minimal, though individuals identifying outside male or female categories showed lower and less variable scores. Age had little impact overall, although younger participants displayed more variability in their responses. Education emerged as a stronger factor—those with advanced degrees demonstrated higher and more consistent affinity scores than those with only some college education. Regional differences were also notable, with respondents from Asia showing higher support for AI in healthcare compared to North America and other regions.
While participants generally viewed digital health positively, there was more caution towards AI integration specifically. Older adults leaned more favourably towards digital tools, while younger respondents were slightly more open to AI itself. These findings highlight the importance of recognising how education and regional context shape public sentiment about emerging technologies in care settings.
Modelling AI Affinity for Predictive Insight
To quantify and predict patient preferences, researchers developed the AI Affinity Score using a weighted system based on responses to a structured survey. Each response was assigned an AI affinity coefficient indicating its level of favourability, and these were aggregated into a score using weights determined by expert assessment. This mathematical model captures a patient’s inclination towards AI within a healthcare context.
The research team trained several machine learning models to predict these affinity scores, using 24 predictors derived from demographic and attitudinal data. Principal Component Analysis (PCA) was used to identify the most relevant features, such as trust in AI, comfort with robotics and familiarity with digital health tools. Models included a deep learning regression model, a classification neural network, a linear regression baseline and a Random Forest classifier.
Interestingly, the linear regression model slightly outperformed the deep learning approach in terms of predictive accuracy, a result attributed to the relatively small dataset. All models showed strong alignment between predicted and observed scores, confirmed by statistical tests. While the deep learning classifier struggled with imbalanced class distribution—especially underperforming in predicting low-affinity categories—adjustments such as class weighting partially mitigated this issue. The Random Forest classifier performed well in medium and high categories but faced similar challenges with low-affinity predictions.
Implications for Care Design and Health Equity
The AI Affinity Score provides a practical tool for personalising healthcare delivery according to individual preferences for AI involvement. When embedded into health systems, this score can inform how much automation is appropriate per patient, enhancing satisfaction and engagement. Those with higher affinity scores might benefit from AI-driven diagnostics or virtual assistants, whereas lower-affinity patients may prefer more traditional, human-centred approaches.
From a policy and planning perspective, understanding the distribution of affinity scores across populations enables targeted resource allocation. For example, regions or communities with high affinity might be early adopters of AI tools, while others may require more education and engagement. Importantly, the model respects patient autonomy and avoids marginalising those who are less comfortable with technology. Categorising patients into functional clusters based on their affinity scores allows for more nuanced segmentation and strategic intervention, with Bayesian models suggesting five distinct levels of AI receptiveness.
Must Read: Rethinking Patient Experience in the Digital Era
Despite its utility, the model has limitations, particularly those common to survey-based research such as selection bias. The study population was skewed towards tech-savvy individuals, as it relied on online data collection. Furthermore, the current model only includes a limited set of demographic variables. Including more granular social determinants of health, geographical factors and behavioural data could improve predictive precision. Nonetheless, the simplicity of the model aids in generalisation and prevents overfitting.
The development of the AI Affinity Score introduces a novel framework for tailoring AI integration in healthcare to individual patient preferences. By capturing attitudes through a mathematically grounded, machine learning-supported model, healthcare systems can make informed decisions about the deployment of technology that respects diversity in acceptance and comfort. Education and regional context appear to be key determinants of affinity, highlighting the need for customised communication and implementation strategies. While further refinement and broader datasets are required, this approach offers a promising path toward truly patient-centred AI adoption in healthcare.
Source: Scientific Reports
Image Credit: iStock
