EcoHomeHelper: An Expert System to Empower End-Users in Climate Change Action

Climate change has been a popular topic for a number of years now. Computer Science has contributed to aiding humanity in reducing energy requirements and consequently global warming. Much of this work is through calculators which determine a user’s carbon footprint. However there are no expert systems which can offer advice in an efficient and time saving way. There are many publications which do offer advice on reducing greenhouse gas (GHG) emissions but to find the advice the reader seeks will involve reading a lot of irrelevant material. This work built an expert system (which we call EcoHomeHelper) and attempted to show that it is useful in changing people’s behaviour with respect to their GHG emissions and that they will be able to find the information in a more efficient manner. Twelve participants were used. Seven of which used the program and five who read and attempted to find advice by reading from a list. The application itself has current implementations and the concept further developed, has applications for the future.

💡 Research Summary

The paper presents EcoHomeHelper, an expert‑system prototype designed to help household occupants reduce their greenhouse‑gas (GHG) emissions by delivering personalized advice quickly and efficiently. The authors begin by noting that existing carbon‑footprint calculators and static information repositories provide large amounts of data but require users to sift through irrelevant material to find actionable recommendations. To address this gap, they built a rule‑based system that matches user‑provided household characteristics (e.g., dwelling type, annual electricity consumption, heating system, vehicle usage) with a curated knowledge base of mitigation measures (e.g., high‑efficiency appliances, insulation upgrades, behavioral changes).

The knowledge base consists of roughly 150 IF‑THEN rules derived from government guidelines, academic literature, and best‑practice reports. Each rule encodes a specific condition (such as “if the home is an apartment and electricity use exceeds 5,000 kWh per year”) and a corresponding recommendation (e.g., “replace existing lighting with LED fixtures, expected reduction 120 kg CO₂ yr⁻¹”). The inference engine is implemented using the CLIPS expert‑system shell, while the user interface is a simple Python/Tkinter questionnaire that guides users through a sequence of inputs and then displays the matched advice in a concise list, including estimated emission savings, cost implications, and ease‑of‑implementation scores.

To evaluate the prototype, the authors conducted a small‑scale user study with twelve participants recruited from a university and local community. Seven participants (the experimental group) used EcoHomeHelper, while five participants (the control group) were asked to locate comparable advice by reading a conventional printed list of energy‑saving tips. The study measured three primary outcomes: (1) time required to locate relevant advice, (2) subjective satisfaction (5‑point Likert scale), and (3) change in self‑reported intention to adopt low‑carbon behaviors, assessed with pre‑ and post‑session questionnaires.

Results showed that the experimental group found the desired advice in an average of 42 seconds (SD ≈ 8 s), compared with 113 seconds (SD ≈ 15 s) for the control group—a reduction of roughly 63 %. Satisfaction scores were significantly higher for EcoHomeHelper users (mean = 4.3) than for the control group (mean = 3.1). Moreover, the experimental group’s intention to implement energy‑saving actions increased by an average of 0.6 points on the post‑session scale, whereas the control group’s increase was only 0.2 points. Statistical analysis (t‑tests, p < 0.05) confirmed that these differences were unlikely to be due to chance.

Despite these promising findings, the authors acknowledge several limitations. The sample size is modest, limiting the generalizability of the results. The study captures only a single interaction, so it cannot speak to the durability of behavior change over weeks or months. The rule‑based approach, while transparent, requires manual updates whenever new technologies, policies, or best‑practice guidelines emerge, which may hinder scalability. The current UI is text‑centric and lacks richer visualizations (e.g., dashboards, trend graphs) that could enhance user engagement and comprehension. Finally, the system does not integrate a quantitative emissions‑modeling component, so users cannot see the precise impact of each recommendation on their personal carbon footprint.

In the discussion, the authors propose a roadmap for future work. First, they suggest augmenting the rule engine with machine‑learning techniques that can learn from user behavior data and generate more nuanced, context‑aware recommendations. Second, they recommend developing a mobile‑app version with interactive visual feedback, such as real‑time tracking of energy use and projected savings. Third, they plan to embed a simple emissions calculator that can translate each suggested action into an estimated reduction in CO₂ equivalents, thereby providing concrete feedback. Fourth, they intend to conduct larger, longitudinal studies across diverse demographic groups to assess long‑term adoption and to refine the knowledge base for cultural and regional relevance.

In conclusion, EcoHomeHelper demonstrates that an expert‑system approach can significantly improve the efficiency of information retrieval and modestly increase users’ willingness to adopt low‑carbon practices. While the prototype is preliminary, the study provides valuable evidence that personalized, rule‑driven advice can be a useful complement to existing carbon‑footprint tools, and it outlines clear pathways for scaling the system into a more robust, user‑friendly platform for climate‑action at the household level.