Faizan Asad, M.Sc. (Tech.), Circular Material Researcher, Arcada University of Applied Sciences, Department of Energy and Materials Technology
Shaheer Alam, M.Sc. in Artificial Intelligence for Sustainable Societies, Erasmus Mundus Joint Master’s Programme,
From isolated tests to a connected research system
Developing a circular material rarely means improving one variable at a time. Recycled materials can vary between batches. Natural fibres may change moisture uptake, melt flow, stiffness and interfacial adhesion simultaneously. A processing adjustment that improves foam expansion may reduce strength. A formulation that performs well in the laboratory may still be difficult to scale, recycle or justify environmentally.
Circular materials research is therefore better understood as a system of interacting choices: feedstock, formulation, processing, microstructure, performance, manufacturability and end-of-life. Systemic thinking encourages researchers to examine these relationships rather than optimize a single outcome in isolation (Norman 2023).
In this context, system intelligence means being able to recognize these interactions and make informed decisions even when the entire material system cannot be fully predicted or controlled. Artificial intelligence can support this way of working because machine-learning methods are well suited to identifying relationships between multiple variables. Material informatics is already being applied to property prediction, materials design and the connection between composition, processing, structure and performance (Tran et al. 2024; Huo et al. 2025).
Connecting materials, manufacturing and product performance data
Materials research generates information about raw materials, manufacturing methods and the performance of finished products. This information is often analysed separately, making it difficult to see how decisions at one stage affect outcomes at another. AI-assisted analysis can help connect these pieces and identify patterns that are worth investigating. AI can help identify combinations for experimental testing and reveal compromises between product performance, ease of manufacturing and recyclability.
Researchers need to interpret the results and test possible explanations, particularly when datasets are small or cover a limited range of conditions (Xu et al. 2023). AI can help connect and analyse information about materials and how they are processed, but human expertise and experimental testing remain essential for turning these insights into reliable improvements.
Guiding experiments with AI and researcher judgement
One of the most practical uses of AI may be helping researchers decide what experiment to perform next. Experimental materials research consumes materials, additives, energy, machine time and researcher time. Repeating broad trial-and-error campaigns can therefore conflict with the resource-efficiency goals of circular research.
Methods such as active learning and Bayesian optimisation can use existing experimental results and model uncertainty to prioritize informative experiments. A university laboratory does not need a fully autonomous or “self-driving” laboratory to benefit from this principle. A much simpler research cycle can already be valuable:
measure → connect → predict → test → reflect
Researchers first generate reliable experimental data. AI helps identify relationships and points towards promising regions of the experimental space. New experiments test those predictions, and their results are returned to the dataset. Researchers then evaluate whether the suggested solution makes physical and practical sense.
The final step, reflection, is crucial. A model may identify a formulation with excellent predicted stiffness, but the material may require an unrealistic processing temperature, make recycling more difficult or provide only a small performance improvement at a substantial environmental cost. A single score can incorporate several objectives and constraints. The issue is whether it adequately represents the research problem. Assessing this requires recognizing which objectives, constraints and consequences matter.
Green AI for more resource-efficient research
Using AI to support sustainability research also requires attention to AI’s own resource demands. Although AI may help researchers develop resource-efficient materials, the computing systems it relies on consume energy and require physical infrastructure. Its environmental benefits should therefore not be taken for granted.
Green AI addresses this concern by emphasising computational efficiency alongside model performance, rather than treating accuracy as the only measure of success (Schwartz et al. 2020). This means considering the computational resources needed to develop and use a model, as well as the value of its results.
For circular materials research, this principle encourages the use of models whose computational demands are proportionate to the task. Greater complexity does not automatically lead to more reliable or useful findings. Where simpler models provide adequate results, they can help limit computational resource use. Green AI therefore adds an essential consideration to sustainability-oriented research: the resources used to generate knowledge also matter.
From AI tool to system-aware research partner
The strongest role for AI in circular materials research is neither replacing the researcher nor functioning merely as a sophisticated calculator. It can instead be viewed as a system-aware research partner: one capable of comparing many variables, exposing patterns, preserving links between different types of data and challenging the next experimental choice. The researcher remains responsible for defining the problem, checking data quality, interpreting physical mechanisms and deciding whether an apparent optimum is technically, environmentally and socially meaningful. AI may identify a pattern; scientific judgement determines whether that pattern deserves to become knowledge.
This human–AI partnership is particularly relevant in universities of applied sciences, where research often connects laboratories, students, companies and real application environments. AI-assisted materials research can also provide students with experience in combining experimental science with data analysis, sustainability assessment and critical evaluation of digital tools. Used responsibly, AI can strengthen the connections between materials, processing, performance and sustainability. It can help circular material research move from isolated measurements towards an evolving representation of the whole research system. The goal should not be a laboratory that thinks for us. It should be a research process that allows us to see more of the system before deciding how to change it.
Huo, Z., Xie, X. & Tong, R. 2025. Machine Learning for Developing Sustainable Polymers. Chemistry – A European Journal, 31, e202500718. DOI: 10.1002/chem.202500718.
Norman, D. 2023. Design for a Better World: Meaningful, Sustainable, Humanity Centered. The MIT Press, Cambridge, Massachusetts.
Schwartz, R., Dodge, J., Smith, N.A. & Etzioni, O. 2020. Green AI. Communications of the ACM, 63(12), 54–63. DOI: 10.1145/3381831.
Tran, H., Gurnani, R., Kim, C., Pilania, G., Kwon, H.-K., Lively, R.P. et al. 2024. Design of functional and sustainable polymers assisted by artificial intelligence. Nature Reviews Materials, 9, 866–886. DOI: 10.1038/s41578-024-00708-8.
Xu, P., Ji, X., Li, M. et al. 2023. Small data machine learning in materials science. npj Computational Materials, 9, 42. DOI: 10.1038/s41524-023-01000-z.
Circular material research involves interacting choices in raw materials, formulation, processing, performance and end-of-life. Artificial intelligence (AI) can help researchers connect these layers by identifying patterns across experimental, processing, imaging and performance data and by supporting prediction and experimental planning. Its value is especially relevant in recycled materials, composites and foamed materials, where variability and trade-offs are common. However, AI should complement rather than replace experiments and expert judgement. Data quality, interpretability, transparency and computational impacts must therefore remain part of responsible AI-assisted materials research.
Vastaa