INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3350 www.rsisinternational.org
Integrating Artificial Intelligence into Environmental Education: A
Rule-Based Expert System for Hornbill Conservation Awareness
Celine Hew Boon Ling, Eugene How, Soon Boon Ming, Soong Jien Yoong and *Nur Zareen Zulkarnain
Faculty of Artificial Intelligence and Cyber Security, Universiti Teknikal Malaysia Melaka, Hang Tuah
Jaya, Durian Tunggal, Melaka, Malaysia
*Corresponding Author
DOI: https://dx.doi.org/10.47772/IJRISS.2025.910000274
Received: 14 October 2025; Accepted: 21 October 2025; Published: 10 November 2025
ABSTRACT
Hornbills are unique birds inhabiting Southeast Asian tropical rainforests, such as in Malaysia, where they have
an important role in maintaining forest biodiversity while acting as forest regulators and seed dispersers. Their
numbers have, however, plummeted because of deforestation, habitat destruction, illegal poaching and hunting,
among others. Field identification and field observation of species by experts is time-consuming and restricted
because they cannot quantitatively estimate hornbill populations. The growing domain of artificial intelligence
(AI) technologies offers new possibilities for more advanced wildlife tracking and conservation awareness
through automation and smart data management. The article explains the design and implementation of a rule-
based expert system with AI incorporation in environmental education on hornbill species and their conservation
status. A rule-based system integrated with image classification is utilized to categorize ten widely spread species
of hornbills in Malaysia and classify them under the respective International Union for Conservation of Nature
(IUCN) Red List categories. A machine learning framework trained using hornbill images was incorporated into
an easily accessible interface to ensure highest usability and accuracy. The evaluation confirmed that the system
generated consistent recognition results with confidence levels above 90% for all species. Apart from its
technical contribution, the system serves as an interactive learning platform that bridges artificial intelligence
with biodiversity conservation. The system enables users to observe the ways AI models recognize species while
being exposed to ecological and conservation information. This integration of technology and environmental
education engenders critical thinking, interdisciplinarity, and Malaysian wildlife awareness by the public. The
study confirms that AI-driven expert systems are effective means of conservation learning and awareness for
maintaining national and international sustainability goals through technology-facilitated learning.
Keywords: Expert system; Artificial intelligence; Hornbill classification; Biodiversity; Conservation
INTRODUCTION
The use of artificial intelligence (AI) in education has widened the avenues of possibilities through which
learners interact with complex scientific and environmental concepts. AI can be employed as a bridge between
theoretical concepts and real-life application in environmental education so that students are able to perceive and
learn about natural phenomena through data-driven means. With continuous loss of biodiversity globally, the
role of technology in the spread of conservation awareness cannot be overstated. Artificial intelligence systems
present efficient and interactive means of detecting, classifying, and analyzing ecological data with pertinent
learning experiences combining environmental science and computational thinking.
Malaysian tropical rainforests are richly biodiverse with hornbills, which are now symbols of Malaysia’s natural
heritage. Hornbills are significant ecological seed dispersers with functions in forest regeneration and ecological
balance. However, a few hornbill species have experienced significant population loss due to deforestation,
habitat loss, poaching and illicit hunting. In the opinion of the International Union for Conservation of Nature
(IUCN, 2024), many species such as the Helmeted and Rhinoceros Hornbills are presently “Threatened” while
others such as the Oriental Pied Hornbill are “Least Concern” but with long-term threats in their environment.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3351 www.rsisinternational.org
Their loss would not just be causing disruption to forest communities but also diminish Malaysia's biodiversity
inheritance.
Traditional research methodologies of hornbill study by manual observation and photo surveys have been widely
practiced but are often limited by observer bias and logistical limitations. Field identification requires wide
expertise and is difficult to scale in remote forest environments. Besides, visually similar species such as the
Great and Wreathed Hornbills may prove difficult to identify without technical assistance. Such shortcomings
proclaim the need for an intelligent system that can automate the process of classification and improve the
accuracy of identification.
In this study, a rule-based expert system using AI techniques is proposed for identifying species of hornbills in
a more effective way and for improving conservation education. There are three objectives of this study. The
first is to automate hornbill identification using rule-based reasoning that detects main species characteristics
such as casque shape, size, and coloration. The second objective is to improve the efficiency and accuracy of
classification of hornbill species using machine learning for image-based classification over manual observation.
The third objective is to spread data-aware conservation awareness via an easily accessible platform that allows
students, educators, and conservationists to explore how AI assists in biodiversity conservation.
The system’s design combines rule-based reasoning and machine learning classification in a single system. The
rule-based component employs structured reasoning to match species features with known patterns, and the
learned model reads visual information to predict outcomes based on learned image features. This integration
approach allows students to understand symbolic thinking and pattern recognition as building blocks of AI.
Positioning AI in an applied environmental context provides students with a concrete example of how technology
supports problem-solving in authentic settings.
Through artificial intelligence and environmental learning integration, this study demonstrates the ability of
intelligent systems to improve scientific learning based on interactivity and efficiency. The study not only
supports technological development but also helps sustainability goals through ecological literacy and
conservation participation. This study demonstrates that AI can serve both as a scientific tool and a learning
medium, enabling users to understand and help preserve Malaysia’s natural heritage. Furthermore, this initiative
aligns with the United Nations Sustainable Development Goals (SDG), particularly SDG 13 (Climate Action)
and SDG 15 (Life on Land), by promoting environmental education literacy and encouraging proactive
participation in biodiversity conservation through technology-enhanced learning.
LITERATURE REVIEW
Environmental education has evolved significantly with the growing adoption of digital and intelligent
technologies. As sustainability issues become more complex, educators and researchers have sought innovative
approaches to make environmental education more interactive and accessible. Among the emerging
technologies, AI has emerged as one of the most promising tools in this transformation, offering new ways to
analyze data, visualize ecosystems, and encourage public participation in conservation efforts. At the same time,
understanding the ecological challenges facing endangered species remains essential for meaningful
environmental learning. This section reviews recent studies on the integration of AI in environmental education
and examines the major environmental threats contributing to the decline of hornbill populations in Malaysia
and the wider Asian region.
Artificial Intelligence in Environmental Education
The integration of AI tools has reshaped the delivery of environmental education by enabling students and the
public to explore complex environmental concepts and theories through simulation, visualization, and intelligent
feedback. Adaptive learning systems, chatbots, and virtual simulations help learners visualize environmental
processes that are often inaccessible in real life (Arif et al., 2025). By processing large amounts of environmental
data and providing adaptive responses, AI supports the shift from passive instruction to active, inquiry-based
learning. It is now possible for students and learners to analyse environmental problems, predict outcomes, and
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3352 www.rsisinternational.org
propose solutions through interactive platforms that transform abstract environmental and ecological information
into engaging, data-driven experiences (Das et al., 2024; Lowan-Trudeau, 2023).
Existing Works
AI applications in environmental learning range from adaptive learning systems and virtual assistants to image-
recognition tools that classify species or analyse ecological data. Machine learning models, particularly those
used in classification and prediction, are increasingly incorporated into environmental science curricula to make
learning more experiential. Konya and Nematzadeh (2024) reviewed recent AI applications in environmental
science and engineering, showing that classification, pattern recognition, and predictive modelling are among
the most effective tools for tasks such as biodiversity monitoring and pollution assessment. Their work shows
that AI can transform how environmental data is processed and used for decision making. Similarly, Sachyani
(2025) demonstrated that integrating AI into environmental education through creative tools such as digital
comics and visualization activities improved students’ engagement and comprehension of ecological
interactions. An article by Rrustemaj (2025) describes how techniques such as virtual tutoring, automated
assessment, and adaptive content delivery has been used effectively across domains. The same techniques can
also be adapted for environmental education, supporting interactive and personalized learning.
Benefit and Challenges
Integrating artificial intelligence into environmental education enhances personalization, interactivity, and
accessibility in learning. Adaptive systems tailor content to individual learners, while AI-driven simulations
allow students to analyse authentic environmental data and visualize complex ecological interactions. Studies
show that AI integration supports engagement and deep learning when aligned with clear instructional goals
(Wang et al., 2024; Ifenthaler et al., 2024). In ecological contexts, classification and predictive models improve
understanding of biodiversity and environmental monitoring (Konya et al., 2024). However, challenges persist
regarding data quality, algorithmic transparency, and equitable access. Poorly curated datasets may bias model
outcomes, while limited computational infrastructure can hinder implementation in resource-constrained
settings. Ethical concerns about data privacy and the risk of overreliance on AI tools further emphasize the need
for human oversight and pedagogical alignment (Rrustemaj, 2025; Zhai et al., 2024).
Environmental Threats to the Hornbill Population
Hornbills play an important ecological role as seed dispersers that help maintain forest balance. However, their
populations in Asia, including Malaysia, are declining due to increasing human activities. These birds depend
on mature forests for nesting and feeding, making them highly vulnerable to environmental disturbances. The
main factors threatening their survival include deforestation, habitat loss, poaching, and illegal hunting (BirdLife
International, 2022).
Deforestation and Habitat Loss
The hornbill populations have declined rapidly as their natural habitats continue to be threatened by deforestation
and extensive logging causing destruction to their habitats. The loss of large hollow trees, which serves as
essential nesting sites, further exacerbates this decline (Misni et al., 2017). As hornbills are large-sized cavity
nesters, the availability of trees with substantial girts and large cavities are crucial for providing suitable nesting
sites to sustain their populations (Kaur, 2020; Misni et al., 2017). In Peninsular Malaysia, over 3.7 million
hectares of forest loss were reported (Aik & Perumal, 2017), affecting many bio-diversities and wildlife
(Abdullah & Hezri, 2008). Planned developments are also one of the environmental threats endangering the
hornbill population (Alamgir et al., 2020) particularly in Sarawak, Malaysia where extensive areas of its intact
forests have been lost in recent decades (Jaafar et al., 2020; Gaveau et al., 2014) thus calling for urgent
conservation efforts by the Sarawakian government.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3353 www.rsisinternational.org
Poaching and Illegal Hunting
The severity of poaching poses significant threats to the hornbill population particularly the Helmeted Hornbill
also known by its scientific name as Rhinoplax vigil, which is currently under the threat of trade for its bright-
colored casque (Beastall et al., 2016; Collar, 2015; ) that is a material often sought after for carving ornamental
items (Beastall et al., 2016). In 2015, the Helmeted Hornbill was uplisted from ‘Near Threatened’ to ‘Critically
Endangered’ on the International Union for Conservation of Nature (IUCN) Red List of Threatened Species due
to these excessive poaching activities (Aik & Perumal, 2017).
METHODOLOGY
This study adopts a design and development methodology to develop an AI-based expert system to support
environmental education and wildlife conservation awareness. This study focuses on ten species of Malaysian
hornbills, applying rule-based reasoning as well as classification principles to support the automation of species
identification and ease in learning for end-users. The methodology includes four major components which are
system design, rule-based reasoning and classification, system architecture, and implementation.
System Design
The system was conceptualized to be a learning and conservation support tool for users, from students to
educators and conservationists, to identify hornbill species and their ecological relevance. The design began with
target species determination and selection of visual and ecological features that could be employed to distinguish
between them. They included body size, casque shape, coloration, and habitat. Information was obtained from
publicly accessible biodiversity databases and field surveys reported in the literature.
A knowledge base was then built with factual data on ten common hornbill species found in Malaysia: Helmeted,
Rhinoceros, Wrinkled, Oriental Pied, Black, Plain-pouched, White-crowned, Bushy-crested, Great, and
Wreathed Hornbills. For each species, corresponding characteristics and conservation status categories were
defined in terms of the International Union for Conservation of Nature (IUCN, 2024). These facts formed the
foundation of the reasoning mechanism of the expert system.
Target User
The expert system is mainly educational, whereby pupils, educators, and the public can access information on
hornbill species via an interactive interface. Users can identify different hornbill species and determine whether
each has been listed as “Threatened”, “Near Threatened”, or “Least Concern” through image uploading. The
same system can also aid forest rangers, conservation organizations, and researchers by allowing quick and
correct hornbill species identification in the field. By fulfilling this dual function, the system helps to promote
both environmental education and effective conservation awareness.
Rule-Based Reasoning and Classification
The expert system uses a classification-centered approach in an AI environment using rule-based reasoning for
hornbill species identification against known visual and ecological attributes. A system of “if-then” rules was
created that relied on such species characteristics as body weight, casque shape, coloration scheme, and
environment. For instance, if the bird is very large with a heavy red casque and occurs in primary forests at low
elevations, then the system diagnoses it as a Helmeted Hornbill, a threatened species. Each hornbill species is
embodied in a unique set of features that collectively form the knowledge base of the system.
The inference engine operates in a forward-chaining methodology beginning from provided attributes and
repeatedly moving through the rule base until it reaches the point where it determines a condition has been
fulfilled. When a rule is fired, the system produces an output indicating the determined species and its
conservation category, i.e., “Threatened”, “Near Threatened”, or “Least Concern”. This rule-based reasoning
technique is equivalent to classification system logic commonly utilized in machine learning, where there is an
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3354 www.rsisinternational.org
association of defined features with specific output classes. Table 1 below presents 13 rules used in the expert
system to perform forward-chaining inferencing.
Table 1. Rule set used in the expert system.
Rule 1 IF the hornbill is very large in size AND has a heavy red casque AND prefers lowland primary
forests, THEN the species is Helmeted Hornbill.
Rule 2 IF the hornbill is large in size AND has a yellow casque curving upward AND is found in lowland
to hill rainforests, THEN the species is Rhinoceros Hornbill.
Rule 3 IF the hornbill is medium in size AND has a wrinkled-looking casque AND lives in lowland forests
of Borneo, THEN the species is Wrinkled Hornbill.
Rule 4 IF the hornbill is medium in size AND has a black and white body with a small casque in yellow or
white AND is active in open and secondary forests, THEN the species is Oriental Pied Hornbill.
Rule 5 IF the hornbill is medium to large AND has a black beak with a simple casque AND prefers lowland
dipterocarp forests, THEN the species is Black Hornbill.
Rule 6 IF the hornbill is large in size AND has a long, narrow casque and no distinctive dark bar on its
throat pouch AND is found in hill and lowland forests of northern Peninsular Malaysia or Thailand,
THEN the species is Plain-pouched Hornbill.
Rule 7 IF the hornbill is large in size AND has a distinct white crown and shaggy crest AND lives in
lowland to hill forests or plantations, THEN the species is White-crowned Hornbill.
Rule 8 IF the hornbill is small to medium AND has a bushy crest AND lives in lowland or secondary
rainforest, THEN the species is Bushy-crested Hornbill.
Rule 9 IF the hornbill is very large in size AND has a bright yellow casque with black markings AND
prefers mature forest tracts, THEN the species is Great Hornbill.
Rule 10 IF the hornbill is large in size AND has a curved casque, black body, white tail and has a distinctive
dark bar on its throat pouch AND lives in evergreen forests up to mountain regions, THEN the
species is Wreathed Hornbill.
Rule 11 If the hornbill species is Helmeted Hornbill OR Rhinoceros Hornbill OR Wrinkled Hornbill OR
Plain-pouched Hornbill OR Great Hornbill OR Wreathed Hornbill THEN category is Threatened.
Rule 12 If the hornbill species is Oriental Pied Hornbill
THEN category is Least Concern.
Rule 13 If the hornbill species is Black Hornbill OR White-crowned Hornbill OR Bushy-crested Hornbill
THEN category is Near Threatened.
The system integrates both rule-based reasoning and machine learning to enhance hornbill species identification
and understanding. The rule-based component extracts key features and characteristics such as body size, casque
color and size, and habitat from hornbill images. It then guides users in interpreting these traits to classify the
species accurately. To complement this, Google’s Teachable Machine was used to train an image classification
model on a curated dataset of hornbill images. The model learns to recognize visual patterns and distinguishing
features of various hornbill species with different behaviors and positions. Once trained, it is deployed within
the system to perform automated species classification.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3355 www.rsisinternational.org
While the machine learning component predicts or classifies the hornbill species, the rule-based system explains
and verifies the classification based on predefined expert knowledge. This combination balances analytical
accuracy with educational value, allowing users to see how expert systems and machine learning work together
to deepen understanding of biodiversity through artificial intelligence.
System Architecture
Hornbill expert system architecture was designed hybrid which combines rule-based reasoning and image-based
classification. With such architecture, the system both acts as a conservation support system and as a learning
tool. Five important components of its architecture are (i) user interface, (ii) knowledge base, (iii) inference
engine, (iv) integration of the trained model, and (v) output module.
The user interface serves as the point of interaction between users and the system. It is a simple interactive stage
where users can input hornbill images that will be used by the system to identify features like body size, casque
shape, color, and type of habitat. The interface was designed with Python in mind to make it accessible to both
learners and teachers. Its easy-to-use interface allows students and conservationists to use the system without
any trouble in leveraging AI concepts through learning biodiversity.
The knowledge base retains information in structured form and rule sets for ten hornbill species in Malaysia that
are Black Hornbill, Bushy-crested Hornbill, Great Hornbill, Helmeted Hornbill, Wreathed Hornbill, White-
crowned Hornbill, Rhinoceros Hornbill, Plain-pouched Hornbill, Oriental Pied Hornbill, and Wrinkled Hornbill.
Each rule sets out the combination of hornbill features and the corresponding species name. For instance, a large
body size with an upcurved yellow casque and a habitat in rainforest lowlands corresponds to the Rhinoceros
Hornbill. All the conservation categories were projected using the International Union for Conservation of
Nature (IUCN, 2024) data but condensed into three kinds that are “Threatened”, “Near Threatened”, or “Least
Concern”. Knowledge base provides the facts based on the inference engine and ensures the accuracy of the
educational material in the system.
The inference engine uses the user input through forward-chaining reasoning. The engine compares the provided
attributes against the knowledge base rules until a matching condition is met. When a rule is triggered, the engine
produces an output consisting of the species' name and its conservation status. This logical operation
demonstrates to students how expert systems mimic human decision-making through structured reasoning.
Apart from the rule-based reasoning system, the system features a classification model that has been trained
using Google Teachable Machine. The system utilized a small image dataset of various species of hornbills to
train the model to learn significant visual patterns. The trained model was then exported and imported into the
Python environment to assist the inference engine in classification tasks. Once the image is uploaded by the user,
the model estimates the probable species, and this is checked against the rule-based result to make it more
accurate. This integration allows the user to observe how the machine learning complements expert systems,
thereby maintaining the application functional as well as educational.
The output module displays the final classification result. It displays the hornbill species name, prediction
confidence level and its conservation category. The interface provides real-time visual and textual feedback to
enable learners to associate observed traits with environmental awareness. The module unites the system’s
double aim of promoting environmental awareness and demonstrating the applied nature of AI in environmental
education.
System Implementation
Expert system implementation was focused on translating conceptual design into an operational framework
combining rule-based thinking and image-based classification. The software development was made to achieve
technical correctness and pedagogic feasibility in order that the system may be employed as an effective species
identification tool and as an educational utility for understanding artificial intelligence applications for
environmental preservation.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3356 www.rsisinternational.org
The system was coded using the Python programming language within the Anaconda environment, where
package management and bundling of multiple components at once were supported. The development and
testing were accomplished through Visual Studio Code. Python was used because of its versatility and simplicity,
making it easy for students and instructors to easily explore how artificial intelligence principles are applied in
real-world applications. The whole process of development had a design-test-refine loop to guarantee stability
and clarity of user interaction.
Figure 1. Model training using Google Teachable Machine.
Google Teachable Machine was used to train a simple image classification model from a carefully curated dataset
consisting of ten hornbill species. The Teachable Machine framework internally applies a convolutional neural
network (CNN) for image training and recognition, allowing simple deployment without complex coding. The
images were each tagged with the species’ names, and the model was trained to recognize distinctive visual
patterns for each of the species. For each species category of the hornbills, at least 50 images of the species were
used to train the model. The images for training were sourced from publicly available Google Images results and
were carefully placed in different folders based on their species. The number of images used in training each
species category was made close to one another to guarantee that the trained model is neutral. The trained model
was then exported in TensorFlow format and placed within the Python environment to allow automated
identification of species. It supports the rule-based reasoning component by increasing the reliability of
classification.
The model was tested and validated on image-based inputs. A 70-30 ratio was used in training and testing the
model. Various images of hornbills with different species were uploaded and processed using the trained model.
Results were compared to authentic species information from the International Union for Conservation of Nature
(IUCN, 2024) to confirm the reliability of both reasoning processes. The results showed that the system was able
to classify hornbill species with consistent confidence levels and provide the appropriate conservation categories,
aligning with IUCN data. The interface is kept minimal such that users can upload a picture and instantly get the
identified species and its conservation status. The lightweight design enables the system to run on standard
computers without any additional installation procedures, making it ready to be used for demonstrations in
classrooms, lab exercises, or public awareness programs.
The deployment realized an operational AI-driven expert system capable of recognizing hornbill species and
providing effective learning experiences. The integration of rule-based reasoning and image-based classification
not only enhanced the accuracy of identification but also demonstrated the potential of technology in facilitating
environmental education. Students can explore artificial intelligence ideas and biodiversity issues at the same
time on a single interactive site using this system. Beyond its technical capabilities, the deployment also
highlights the broader role of AI in supporting sustainable development. By enabling users to understand species
diversity, conservation status, and the consequences of habitat loss, the system cultivates environmental
responsibility and critical thinking among learners. It empowers educators to incorporate real-world data and AI
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3357 www.rsisinternational.org
applications into their lessons, thereby bridging computational literacy with ecological awareness. As a result,
this system directly supports SDG 13 (Climate Action) and SDG 15 (Life on Land).
RESULTS AND DISCUSSION
The hornbill image classification system was designed to automatically recognize ten hornbill species that are
found in Malaysia which are the Black Hornbill, Bushy-crested Hornbill, Great Hornbill, Helmeted Hornbill,
Wreathed Hornbill, White-crowned Hornbill, Rhinoceros Hornbill, Plain-pouched Hornbill, Oriental Pied
Hornbill, and Wrinkled Hornbill. The back-end classification model utilized a machine learning model that was
trained using a standard classification approach and then implemented in a graphical user interface (GUI) for
verification and user convenience. The GUI was intended to facilitate users in uploading an image, visualizing
the results of the predictions, and seeing the conservation status of the species that were detected. Apart from
being a testing site, the interface also acts as an educator that shows how artificial intelligence can be used in
wildlife identification and learning diversity.
System Performance
The system was tested with hornbill images of varying orientations, lightings, and backgrounds to test the
stability of model recognition. Figure 2 shows a sample result for a sample image of Black Hornbill. The model
correctly predicted the species with 99.29% confidence level and displayed its respective conservation status as
“Near Threatened” which is correct according to the International Union for Conservation of Nature (IUCN,
2024). The GUI visualization includes a horizontal confidence bar that represents the probability distribution of
all possible species. This enables users to visualize the confidence of the model and, therefore, make the
classification more transparent. The unequivocal dominance of the top prediction, reflected in the confidence
plot, indicates that the machine learning model constructed an explicit decision boundary in the feature space
that well is discriminated against the correct class from the remaining nine species.
The same test procedure was extended to the other hornbill species to validate the consistency of the model
across different samples. Figure 3 shows additional examples of the classification interface, which consistently
provide correct identifications of species along with correct conservation labels. The interface is of high
confidence even when images vary in light and background, displaying the strength of the model. GUI design is
maintained as simple and informative, with every prediction displayed along with its percentage confidence
level, image, and conservation category. This simple explanation does not only optimize usability but also
connects the technical outcome to its ecological basis, allowing users to understand both how AI classification
works and why hornbill conservation matters.
Figure 2. Detailed confidence level for specific class prediction.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3358 www.rsisinternational.org
Figure 3. Examples of the classification interface that includes three threats level.
Table 2 summarizes the classification results of model testing for all ten species. The classifier maintained the
maximum levels of prediction confidence, mostly over 90%, on approximately 20 to 30 unseen test images. The
Helmeted Hornbills and the Rhinoceros were identified perfectly with highly confident levels due to their regular
casque shape and large body size. The Oriental Pied and White-crowned Hornbills were also identified
accurately based on their pale coloration and facial coloration. Minor confusion occurred between the Great and
Wreathed Hornbills, whose superimposing colors and resembling casque shapes often lead to visual ambiguity
even to field observers. Such misclassifications are unavoidable in fine-grained image classification problems
where inter-species difference is fine.
Table 2. Classification outcomes of model testing on ten different hornbill species.
No. Hornbill Species Conservation Status Observed Prediction
Confidence Level (%)*
Classification Outcome
1. Black Near Threatened 99.29 Near Threatened
2. Bushy-crested Near Threatened 99.53 Near Threatened
3. Great Threatened 96.95 Threatened
4. Helmeted Threatened 98.14 Threatened
5. Wreathed Threatened 96.31 Threatened
6. White-crowned Near Threatened 98.9 Near Threatened
7. Rhinoceros Threatened 99.34 Threatened
8. Plain-pouched Threatened 99.01 Threatened
9. Oriental Pied Least Concern 99.92 Least Concern
10. Wrinkled Threatened 97.76 Threatened
* Confidence level values were derived from the model's graphical output at test run and represent average
observed levels across sample runs.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3359 www.rsisinternational.org
The findings in Table 1 and Figures 2 and 3 consistently indicate that the expert system is very accurate and
robust in identifying species. The classifier consistently produced confidence levels higher than 90%, consistent
with the validation accuracy attained at training. These findings validate that the machine learning algorithm
generalized to new data effectively, irrespective of variations in lighting or background. The qualitative findings
also validate that morphological characteristics such as casque shape, beak curvature, and plumage pattern are
robust distinguishing features for hornbill identification. The minor ambiguity between visually confused species
is anticipated and strengthens documented field-identification difficulties and underscores the capability of AI
to support human professionals in reducing such mistakes.
Although this analysis was based on manual trials instead of statistical benchmarking, the outcomes provide
strong evidence of the model’s reliability under realistic conditions. Interactive visualization of confidence
scores by the GUI improves interpretability and encourages trust in the predictions of the system. By allowing
users to experience prediction confidence directly, the system introduces an implicit learning dimension
regarding how artificial intelligence manages uncertainty and confidence estimation.
Education and Conservation Impact
Aside from technical accuracy, the educational and social impacts of the expert system are large. The GUI turns
what would otherwise be a computation-only system into an experiential learning platform. Not only can users
test hornbill images to determine their species but also regarding the conservation status of their species. This
approach fosters environmental education objectives by placing scientific facts into a fun and interactive context.
The integration of IUCN Red List categories such as “Threatened”, “Near Threatened”, or “Least Concern”
bridges AI with environmental literacy, enabling users to connect technology with real-world sustainability
issues.
For students, the system teaches how AI can be employed to solve environmental issues and how digital skills
can be learned. The system offers cross-disciplinary learning opportunities where computer science and
environmental science intersect with sustainability. Students get to learn key concepts in AI such as feature
extraction and classification from real-world conservation data. For conservation professionals, the model's
strong recognition capability provides a lightweight tool that can contribute towards preliminary species
identification when conducting field surveys, particularly in regions with limited expert presence. Government
agencies and wildlife organizations can also adapt this design to fit public-awareness programs or citizen-science
programs focused on hornbill monitoring.
CONCLUSION
The hornbill expert system effectively proves AI applications’ capability to enrich environmental education and
wildlife awareness. By combining rule-based reasoning with image-based classification, the system provides an
engaging and interactive learning experience for hornbill species identification while communicating their
conservation status. Integration of a machine learning classifier into an intuitive interface enables students to
experience how computer vision processes images and make decisions, bridging the gap between technological
and environmental knowledge. The test outcomes revealed that the system works with very high accuracy and
stability for all ten species of hornbills with prediction confidence above 90%. These findings validate the
system’s value not only as a reliable tool for classification, but also as an educational tool that breaks down
complex AI concepts into simple learning experiences.
This study points to the ways in which AI-driven methods can transform traditional environmental education by
extending experiential learning and enhancing digital literacy. Through interactive classification and
conservation-based comments, students are encouraged to explore the intersection of technology and
sustainability. The system also demonstrates validity in real-world application for forest rangers, scientists, and
conservation groups seeking practical methods of hornbill identification. Its pedagogical nature ensures that
students and the public can meaningfully engage with information pertaining to conservation efforts,
encouraging environmental stewardship through learning technology. By supporting education on biodiversity
and sustainable ecosystems, the system directly contributes to the realization of SDG 15 (Life on Land), while
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3360 www.rsisinternational.org
its integration into environmental education frameworks supports SDG 13 (Climate Action) through increased
awareness and responsible stewardship of natural resources.
Future work can generalize the quantitative analysis of the model with more extensive datasets and performance
measurements in computing such as precision, recall, and F1-score. Further refinement of user interface and
inclusion of visualization explainable-AI functionalities would enhance interpretability and transparency for
learners. The system could also be adapted for web or mobile deployment to support outdoor biodiversity
learning and citizen-science activities. In summary, this work provides a foundation for artificial intelligence
integration into environmental learning and demonstrates how technology might be efficiently utilized to
enhance conservation awareness and sustainability literacy in Malaysia and beyond.
REFERENCES
1. Abdullah, S. A., & Hezri, A. A. (2008). From forest landscape to agricultural landscape in the developing
tropical country of Malaysia: pattern, process, and their significance on policy. Environmental
Management, 42(5), 907-917.
2. Aik, Y. C., & Perumal, B. (2017). Distribution of Hornbills and important Hornbill landscapes setting
site conservation priorities for Peninsular Malaysia. In 7th International Hornbill Conference.
3. Alamgir, M., Campbell, M. J., Sloan, S., Engert, J., Word, J., & Laurance, W. F. (2020). Emerging
challenges for sustainable development and forest conservation in Sarawak, Borneo. PloS one, 15(3),
e0229614.
4. Arif, M., Ismail, A., & Irfan, S. (2025). AI-powered approaches for sustainable environmental education
in the digital age: A study of Chongqing International Kindergarten. International Journal of
Environment, Engineering and Education, 7(1), 35-47.
5. Beastall, C., Shepherd, C. R., Hadiprakarsa, Y., & Martyr, D. (2016). Trade in the Helmeted Hornbill
Rhinoplax vigil: the ‘ivory hornbill’. Bird Conservation International, 26(2), 137-146.
6. BirdLife International. 2022. State of the World’s Birds: 2022 Annual Update. Retrieved from
https://www.birdlife.org/papers-reports/state-of-the-worlds-birds-2022/
7. Collar, N. J. (2015). Helmeted Hornbills Rhinoplax vigil and the ivory trade: the crisis that came out of
nowhere. BirdingASIA, 24, 12-17.
8. Das, S., Anowar, S., & Chakraborty, S. (2024). The integration of AI technology into environmental
education. Life as basic science: An overview and prospects for the future, 1, 223-247.
9. Gaveau, D. L., Sloan, S., Molidena, E., Yaen, H., Sheil, D., Abram, N. K., ... & Meijaard, E. (2014).
Four decades of forest persistence, clearance and logging on Borneo. PloS one, 9(7), e101654.
10. Jaafar, W. S. W. M., Said, N. F. S., Abdul Maulud, K. N., Uning, R., Latif, M. T., Muhmad
Kamarulzaman, A. M., ... & Takriff, M. S. (2020). Carbon emissions from oil palm induced forest and
peatland conversion in sabah and Sarawak, Malaysia. Forests, 11(12), 1285.
11. Kaur, R., Ramli, R., Ancrenaz, M., Hassan, H. E. L. S. O. N., Ahmad, E., Ratag, M., ... & Rajak, A.
(2020). Estimating the availability of potential hornbill nest trees in a secondary forest patch in
Kinabatangan, Sabah. Forktail, 36(1), 56-62.
12. Konya, A., & Nematzadeh, P. (2024). Recent applications of AI to environmental disciplines: A review.
Science of The Total Environment, 906, 167705.
13. Lowan-Trudeau, G. (2023). Digital technologies and environmental education. The Journal of
Environmental Education, 54(1), 1-7.
14. Misni, A., Rasam, A. R. A., & Buyadi, S. N. A. (2017). Spatial analysis of habitat conservation for
hornbills: a case study of Royal Belum-Temengor forest complex in Perak State Park, Malaysia.
Pertanika Journal of Social Science and Humanities, 25, 11-20.
15. Naniwadekar, R., Rathore, A., Shukla, U., Chaplod, S., & Datta, A. (2019). How far do Asian forest
hornbills disperse seeds?. Acta Oecologica, 101, 103482.
16. Norouzzadeh, M. S., Nguyen, A., Kosmala, M., Swanson, A., Palmer, M. S., Packer, C., & Clune, J.
(2018). Automatically identifying, counting, and describing wild animals in camera-trap images with
deep learning. Proceedings of the National Academy of Sciences, 115(25), E5716-E5725.
17. Rrustemaj, M. (2025). Artificial Intelligence (AI) in Education. International Journal of Research and
Innovation in Social Science, 9(5), 867-876.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 3361 www.rsisinternational.org
18. Sachyani, D., & Gal, A. (2025). Artificial Intelligence Tools in Environmental Education: Facilitating
Creative Learning about Complex Interaction in Nature. European Journal of Educational Research,
14(2).
19. The IUCN Red List of Threatened Species (no date) IUCN Red List of Threatened Species. Available
at: https://www.iucnredlist.org/ (Accessed: 2 October 2025).
20. Ifenthaler, D., Majumdar, R., Gorissen, P., Judge, M., M., Mishra, S., Raffaghelli, J., & Shimada, A.
(2024). Artificial intelligence in education: Implications for policymakers, researchers, and practitioners.
Technology, Knowledge and Learning, 29(4), 1693-1710.
21. Wang, S., Wang, F., Zhu, Z., Wang, J., Tran, T., & Du, Z. (2024). Artificial intelligence in education: A
systematic literature review. Expert Systems with Applications, 252, 124167.
22. Zhai, C., Wibowo, S., & Li, L. D. (2024). The effects of over-reliance on AI dialogue systems on students'
cognitive abilities: a systematic review. Smart Learning Environments, 11(1), 28.