top of page
g80_edited.png
Writer's pictureElina Halonen

What are the biggest challenges of using ontologies in behavioural science?

In the previous post, we explored how ontologies provide structure, standardisation, and collaboration in behavioural science. They help us organise complex data and facilitate interdisciplinary research but ontologies aren’t perfect. In fact, they face significant challenges when it comes to representing the full complexity of human behaviour.


Human behaviour is messy, dynamic, and shaped by an intricate web of factors—social, cultural, biological, psychological—and ontologies, by nature, have to simplify that web. But does this simplification help or hinder us? Are we losing valuable nuance in the process? And how well do ontologies integrate the diverse and sometimes competing theories in behavioural science?


In this post, we’ll dive into these questions and explore the limitations of using ontologies in a field as complex as behavioural science.


Why is it so hard to represent complex human behaviour with ontologies?

Ontologies are, at their core, simplifications. They reduce complex phenomena into neatly categorised concepts and relationships, which makes data easier to analyse and interpret. But here’s the problem: human behaviour is rarely neat or linear. It’s influenced by countless variables—many of which are non-linear, contextual, and constantly changing.


Let's take decision-making as an example: it might be influenced by internal factors like motivation, but also by external variables such as social norms, environmental cues, or stress levels. These factors don’t always follow predictable patterns - instead, they interact in dynamic and sometimes contradictory ways. Ontologies tend to create fixed categories which means they can struggle to capture this kind of fluid, multi-layered behaviour.


Think of it this way: ontologies are like maps. They’re useful for getting from point A to point B, but they can’t capture the unpredictability of traffic, weather, or road closures along the way. Similarly, while ontologies provide a structured framework for behaviour, they may miss important contextual nuances or complex causal interactions.


Are we oversimplifying behaviour by using ontologies?

This brings us to the question of oversimplification. By reducing behaviour to fixed categories and relationships, do we risk losing the complexity that makes behaviour so challenging—and interesting—to study?


For instance, take the concept of habits. Habits are automatic, learned behaviours triggered by cues in the environment. But the process of habit formation is far from straightforward. It involves motivation, reinforcement, environmental design, and social influences, among other factors. Can we really capture this complexity with a few static categories like "cue" and "response"?


Ontologies, by necessity, simplify this process, which can be helpful when dealing with large datasets or comparing multiple studies. But we need to be careful. Oversimplification might lead us to overlook important moderating factors like culture, socioeconomic status, or individual differences in personality or cognitive style.

In some cases, an oversimplified ontology might even lead to misleading conclusions. For example, if we categorise behaviours without fully accounting for their complexity, we might design interventions that miss key influencing factors—such as trying to change dietary habits by focusing on individual motivation without addressing food availability or cultural preferences.


What are the practical roadblocks to developing and maintaining ontologies?

Ontologies don’t just face theoretical challenges; there are also practical roadblocks to their development and upkeep.


First, developing a robust ontology is time-intensive and resource-heavy. It requires significant expert input to ensure that all key concepts are accurately defined and categorised. Even with automated tools like there’s no substitute for expert knowledge, especially when it comes to interpreting complex behavioural data.


Second, ontologies need to be constantly updated to reflect new research and evolving theories. Behavioural science is a rapidly changing field, and what we know about behaviour today may shift dramatically in the next few years. Without regular updates, ontologies risk becoming outdated, leading to inaccurate or incomplete representations of behaviour.


Finally, there’s the issue of data integration. Different studies use different methodologies, populations, and intervention designs, which makes it difficult to fit all of this data neatly into a single ontology. Even when we have a shared vocabulary, there’s still the challenge of aligning the underlying assumptions and methodologies that different researchers bring to the table.


The complexity dilemma

Ontologies undoubtedly offer valuable structure and standardisation for behavioural science, but they also come with significant limitations. While they help us organise data and streamline collaboration, they run the risk of oversimplifying the complex, dynamic nature of human behaviour. As we’ve seen, integrating competing theories, addressing non-linear relationships, and keeping ontologies up to date are no small feats.


These limitations don’t mean we should abandon ontologies. Instead, we need to approach them with a critical eye—using them as tools, not catch-all solutions. Ontologies work best when they serve as a starting point, not an endpoint, for understanding behaviour.


In the next post, we’ll explore how ontologies can still be transformative—particularly in designing interventions across diverse fields like public health, business, and technology. We’ll look at real-world applications of ontologies and how they can help scale behaviour change efforts across sectors. Stay tuned for a deeper dive into the practical power of ontologies!


 

Further reading:

  • Beatty, A. S., Kaplan, R. M., & National Academies of Sciences, Engineering, and Medicine. (2022). Understanding Ontologies. In Ontologies in the Behavioral Sciences: Accelerating Research and the Spread of Knowledge. National Academies Press (US). (download)

  • Beatty, A. S., Kaplan, R. M., & National Academies of Sciences, Engineering, and Medicine. (2022). How Ontologies Facilitate Science. In Ontologies in the Behavioral Sciences: Accelerating Research and the Spread of Knowledge. National Academies Press (US). (download)

  • Beatty, A. S., Kaplan, R. M., & National Academies of Sciences, Engineering, and Medicine. (2022). Why Ontologies Matter. In Ontologies in the Behavioral Sciences: Accelerating Research and the Spread of Knowledge. National Academies Press (US). (download)

  • Castro, O., Mair, J. L., von Wangenheim, F., & Kowatsch, T. (2024, February). Taking Behavioral Science to the next Level: Opportunities for the Use of Ontologies to Enable Artificial Intelligence-Driven Evidence Synthesis and Prediction. In BIOSTEC (2) (pp. 671-678). (download)

  • Hastings, J., West, R., Michie, S., Cox, S., & Notley, C. (2022). Ontologies for the Behavioural and Social Sciences: Opportunities and challenges. (download)

  • Larsen, K. R., Michie, S., Hekler, E. B., Gibson, B., Spruijt-Metz, D., Ahern, D., ... & Yi, J. (2017). Behavior change interventions: the potential of ontologies for advancing science and practice. Journal of behavioral medicine, 40, 6-22. (download)

  • Mac Aonghusa, P., & Michie, S. (2020). Artificial intelligence and behavioral science through the looking glass: Challenges for real-world application. Annals of Behavioral Medicine, 54(12), 942-947. (download)

  • Michie, S., West, R., & Hastings, J. (2019). Creating ontological definitions for use in science. Qeios. (download)

  • Norris, E., Finnerty, A., Hastings, J., Stokes, G., & Michie, S. (2019). Identifying and evaluating ontologies related to human behaviour change interventions: a scoping review. (download)

  • Sharp, C., Kaplan, R. M., & Strauman, T. J. (2023). The use of ontologies to accelerate the behavioral sciences: Promises and challenges. Current Directions in Psychological Science, 32(5), 418-426. (download)

Comments


bottom of page