Page 712
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Human-In-The-Loop AI for Precision Agriculture Scoping Review
R.N.I. Basnayake
1
*, G.M.S.C Gajendrasinghe
2
1
Department of Computer and Data Science, NSBM Green University, Sri Lanka
2
Department of Software Engineering and Computer Security, NSBM Green University, Sri Lanka
*
Corresponding Author
DOI:
https://doi.org/10.51583/IJLTEMAS.2026.15020000061
Received: 21 February 2026; Accepted: 27 February 2026; Published: 16 March 2026
ABSTRACT
This scoping study explores the role of Human-in-the-Loop Artificial Intelligence (HITL AI) in precision
agriculture and evaluates the benefits of using human expertise in combination with Artificial Intelligence (AI)
systems in decision-making within modern smart agricultural environments. The development of Artificial
Intelligence, Machine Learning, Internet of Things, and robotics has significantly impacted modern agriculture
by providing automated crop monitoring, disease detection, yield prediction, and smart farm management
systems. However, Artificial Intelligence systems also face challenges in terms of understanding, interpretability,
flexibility, and trustworthiness in modern smart agricultural environments. This study is based on the literature
regarding human-in-the-loop systems, human-centric Artificial Intelligence systems, and collaborative robotics
systems in the context of smart agriculture. The structured scoping study methodology has been followed to
identify and evaluate studies regarding Artificial Intelligence systems in smart agricultural environments, with a
focus on automation-centric Artificial Intelligence systems and human-centric Artificial Intelligence systems
within the context of Agriculture 5.0 concepts. The study concludes that although automation-centric AI systems
show high accuracy in simulated smart agricultural environments, Human-In-The-Loop (HITL) AI systems show
higher robustness in smart agricultural environments. Explainability in AI has shown significant potential in
supporting the effectiveness of HITL AI systems in decision-making within smart agricultural environments.
The study also identifies some important gaps in the literature regarding HITL AI systems in smart agriculture.
The study concludes that for the development of modern smart precision agriculture, collaborative intelligence
within smart agricultural environments is necessary to create sustainable smart agriculture systems.
Keywords: Human-in-the-Loop AI; Precision Agriculture; Explainable Artificial Intelligence; Smart Farming;
Agriculture 5.0
INTRODUCTION
Agriculture currently faces a major revolution in terms of technology, mainly driven by the increased global
demand for food, environmental factors, and the availability of resources. Precision agriculture was developed
as a data-driven approach to address the aforementioned factors, mainly using sensing technologies, Machine
Learning (ML), and automation systems. Previous reviews on ML in agriculture identified the significant use of
data analytics on crop production, soil irrigation, and animal systems [1],[4]. However, there are also broader
conceptual discussion on the agriculture 5.0 concept, highlighting the need for a shift from automation to a more
human-centric and sustainable digital agriculture approach [2],[17]. Recent developments in deep learning
methods show significant improvements in the performance of ML models for crop disease detection and plant
health monitoring using images. The use of Convolutional Neural Networks (CNNs), EfficientNet, and transfer
learning methods shows significant classification performance on standard benchmarks [5],[7]. The EfficientNet
method achieves robust performance in crop disease recognition using a multi-class approach in controlled
environments [6]. In addition, detailed performance analyses show that deep learning methods are more effective
compared to ML methods, especially when the dataset used for training the models is wellbalanced and high
quality [5]. However, the evaluation of ML methods was performed in a controlled environment and does not
Page 713
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
reflect the variations that exist in a real-world environment. However, there are some limitations associated with
fully automated systems, and some of the limitations are as follows: The agricultural environment is highly
dynamic and varies due to climatic variations, heterogeneous soil, and region-wise agricultural practices. Fully
automated systems might not perform well due to the lack of generalization associated with the use of the AI
model, especially when the data set used for training the model is small.
The use of fully automated systems does not guarantee robustness, as discussed in the cyber-physical agricultural
systems research, where the environment is highly uncertain and dynamic [9]. The use of human expertise in the
form of the Human-in-the-Loop(HITL) approach has gained significant interest, and the use of human expertise
in the form of HITL approach has shown significant benefits in the fields of agricultural robotics and
collaborative control systems, where the use of human expertise has improved the efficiency of the allocation
process and the classification process[9],[10]. The use of human expertise in the form of the HITL approach in
the context of human-guided large language model (LLM) pipelines for farm management insights has shown
that human expertise is essential in the context of the agricultural intelligence system, as the use of human
expertise can prevent unreliable recommendations and hallucinations [13]. Parallel to the progress achieved by
human-in-the-loop systems, the concept of explainable artificial intelligence (XAI) has emerged as a research
topic to improve the level of transparency and user trust. Actionable explanation frameworks and feature
attribution techniques are being developed to make AI decisions more interpretable and operationally relevant
to the user [2], [12], [16].
This is particularly important in the context of agriculture, where farmers/agronomists need to understand not
only the predictions, but the explanations as well, before making management decisions. Without the
explainability aspect, even high-performing AI systems may not be adopted due to the lack of understanding of
the decision-making process. In addition to the algorithms, the infrastructure is another important aspect that
affects the implementation of AI systems in agriculture. IoTbased sensing systems, along with various data
modalities, are being used to develop real-time monitoring systems for the environment [4], [14]. However, the
issue of data heterogeneity, the lack of reusable cyberinfrastructure, and machine learning training pipelines is
still a barrier to the widespread implementation of AI systems. While a considerable body of literature has
explored the applications of machine learning techniques in agriculture, relatively few studies have attempted to
synthesize the interplay between precision agriculture, Human-in-the-Loop systems, and Explainable AI within
the context of Agriculture 5.0. While many studies have explored the performance of automation systems,
relatively little research has focused on collaborative approaches to intelligent systems, where both efficiency
and expertise are important considerations.
Figure 1. Agriculture 5.0 concepts [11]
Therefore, the objective of this scoping review is to examine the existing research on the concept of Human-
inthe-Loop AI systems, specifically under the context of precision agriculture. The objectives of the study are as
follows: (1) to explore the development of AI systems in precision agriculture, (2) to evaluate the impact of the
application of the concept of Human-in-the-Loop on the robustness of AI systems, (3) to evaluate the impact of
Page 714
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Explainable AI on the establishment of trust, and (4) to explore the avenues for future research under the context
of human-centric Agriculture 5.0
METHODS
Review Design
The research study utilizes a scoping review methodology to extensively examine the existing literature on the
concept of Human-in-the-Loop Artificial Intelligence in the context of precision agriculture. The reason for
selecting the scoping review methodology is the multidisciplinary and dynamic nature of the field, which
includes areas such as machine learning, agricultural robotics, explainable AI, IoT-based sensor systems, and
cyber-physical infrastructure. The multidisciplinary and dynamic nature of the field makes it suitable for
conducting a scoping review, where the research study is more inclined towards conceptual intersections and
research gaps, rather than conducting statistical analysis. The research study has followed the standard steps of
conducting a scoping review, which includes framing research questions, conducting research, selecting studies,
and synthesizing the literature.
Research Questions
This review is structured along the following four broad research questions. First, the existing applications of
artificial intelligence and machine learning technologies are reviewed in the context of precision agriculture
systems. Second, the inclusion of human-in-the-loop systems is reviewed as a component of artificial intelligence
systems. Third, the place of Explainable Artificial Intelligence (XAI) is reviewed to improve the level of
transparency and trust associated with precision agriculture systems. Lastly, the shortcomings of the existing
literature on human-in-the-loop systems are reviewed as they relate to precision agriculture systems.
Eligibility Criteria
To ensure the relevance and academic credibility of the selected studies, the selection was based on certain
predefined inclusion and exclusion criteria. For example, the inclusion criteria were set as follows: the selected
studies were restricted to peer-reviewed journal articles and conference papers within a specified time frame
from 2020 to 2026. Moreover, the selected studies were expected to focus on artificial intelligence, machine
learning, deep learning, human-in-the-loop computing, explainable AI, IoT-based sensing solutions, robotics,
and cyber-physical systems within an agricultural context. In addition, the selected studies were expected to
focus on certain related themes such as precision agriculture, smart farming, and Agriculture 5.0 frameworks.
Information Sources and Search Strategy
Systemic search strategies on significant scientific databases, including IEEE Xplore, Springer Link,
ScienceDirect, MDPI, Frontiers, and Wiley Online Library, are performed to collect the relevant studies. The
search process includes a combination of keywords related to the research topic, such as “Human-in-the-Loop
AI in Agriculture,” “Precision Agriculture using Machine Learning,” “Explainable AI in Smart Agriculture,”
“Agricultural Robotics and Collaborative Intelligence,” “IoT and ML in Agriculture,” and “Agriculture 5.0,”
using Boolean operators AND and OR to make the search more specific and relevant to the topic, focusing on
the recent developments in AI-based agricultural systems and human-centric intelligence systems.
Study Selection
A total of 860 records were identified across multiple databases, including Google Scholar (n = 450), IEEE
Xplore (n = 120), MDPI (n = 80), SpringerLink (n = 95), ScienceDirect (n = 70), and Frontiers (n = 45). After
removing 210 duplicate records, 650 studies remained for title and abstract screening. Of these, 580 were
excluded due to lack of relevance to Human-in-the-Loop AI in agriculture or absence of AI components. Seventy
full-text articles were assessed for eligibility. Following full-text evaluation, 53 articles were excluded for not
incorporating HITL mechanisms, lacking validation methodology, or not focusing on precision agriculture.
Finally, 17 studies were included in the qualitative synthesis.
Page 715
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Data Charting and Extraction
In the process of data extraction, a structured charting methodology was used to ensure consistency in the
extraction process for the chosen studies. In the context of the studies, relevant information was recorded for
each paper, which included the publication details, the domain of the application, the AI techniques used, the
inclusion of Human-in-the-Loop integration, the inclusion of explainability mechanisms, the performance
metrics used, and the limitations of the studies. The extracted data attributes were further clustered under
thematic groupings for a comparative analysis in the Results section.
Figure 2. PRISMA-SCR flow diagram illustrating the study selection process.
Page 716
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
RESULTS AND SYNTHESIZED FINDINGS
Table 1. Summary of Included Studies on HITL AI in Precision Agriculture
Study
AI Model Type
Validation Method
Field
Deployment
Status
Adaptability
Mechanism
Benos et al.
(2021)
ML, statistical models
Simulation / Dataset
Conceptual /
Analytical
No structured human
feedback
Razak et al.
(2024)
XAI frameworks
Literature-based
review
Conceptual
Human-centered
explainability
Adve et al.
(2025)
Large-scale AI
systems
Mixed (simulation +
collaborative pilots)
Pilot
Collaborative
research feedback
loops
Sudha & Loret
(2026)
ML + IoT integration
Simulation +
Prototype testing
Prototype
Limited adaptive
learning
Ngugi et al.
(2024)
CNN, Deep Learning
Dataset-based
validation
Prototype
No dynamic
feedback
Subhan et al.
(2026)
EfficientNetB0 (DL)
Controlled dataset
Prototype
Static training model
Yakkala et al.
(2025)
Deep Learning
(CNN)
Simulation
Prototype
Early-stage model
updates
Dolatabadian et
al. (2025)
ML image
classification
Dataset validation
Prototype
No HITL mechanism
Sreeram & Nof
(2021)
Cyber-physical HITL
system
Conceptual +
simulation
Pilot
Human decision
integration
Deka et al.
(2025)
HITL planning &
control AI
Simulation +
experimental
Pilot
Iterative human
feedback loop
Benos et al.
(2025)
Explainable AI +
HAR models
Experimental
validation
Prototype
SHAP-based
interpretability
Saranti et al.
(2025)
Actionable XAI
(AxAI)
Experimental
Prototype
Adaptive
classification
feedback
Mourtzinis et al.
(2025)
LLM + HITL pipeline
Real-world case
analysis
Pilot
Human validation
before deployment
Study
AI Model Type
Validation Method
Field
Deployment
Status
Adaptability
Mechanism
Emon et al.
(2025)
IoT + ML integration
Simulation + field
testing
Prototype
Limited human
override
Waltz et al.
(2025)
ML
cyberinfrastructure
Conceptual + pilot
Pilot
Infrastructure-level
adaptability
Arafat (2025)
Explainable DL
(Hyperspectral)
Dataset validation
Prototype
Human-guided
segmentation
Holzinger et al.
(2025)
Human-centered AI
framework
Conceptual
Conceptual
Strong human-in-
theloop design
Page 717
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
The reviewed studies reveal a progressive shift in precision agriculture from fully automated AI systems toward
more collaborative and human-centric architectures. Four dominant thematic clusters emerge: (1) deep
learningbased crop disease detection, (2) IoT-assisted environmental monitoring systems, (3) Human-in-the-
Loop (HITL) planning and control architectures, and (4) explainable AI (XAI) mechanisms supporting trust and
transparency.
Deep learning approaches, particularly CNN, EfficientNet, and transfer learning models [5], [8], demonstrate
high classification accuracy in crop disease detection tasks.
However, most of these systems are validated in controlled or dataset-driven environments, limiting their
adaptability to dynamic field conditions. IoT-assisted systems integrate sensing and machine learning for
irrigation management, yield estimation, and environmental monitoring [1], [4], [14], primarily emphasizing
automation efficiency.
In contrast, HITL-based architectures [9], [10], [13] incorporate expert feedback into planning and control
processes, improving robustness under uncertainty. Explainable AI frameworks [2], [11], [12], [16] further
enhance system transparency, supporting human understanding and decision confidence.
Despite this shift toward collaborative intelligence, the reviewed studies indicate limited integration between
automation, adaptability, and explainability mechanisms, highlighting the need for more cohesive Human-inthe-
Loop architectures in Agriculture 5.0 systems.
DISCUSSION
The synthesized results show that, on one hand, precision agriculture has witnessed rapid advancements through
automation-driven artificial intelligence systems, yet the journey towards collaborative intelligence remains a
work in progress.
While the crop disease detection using deep learning and IoT-based systems show significant predictive accuracy
and operational efficiency, as shown in [5],[8] and [14], the systems are also seen to operate on a fixed dataset
and show poor contextual adaptability.
As the agricultural environment is highly dynamic, the automation-driven systems might face difficulty in
generalizing the data and the decision-making process, as shown in [14].
The introduction of HITL systems also show promise in addressing some of the issues, as the systems are
designed to incorporate human expertise in the decision-making process of the automation-driven systems, as
shown in [9] and [10]. Moreover, the use of explainable artificial intelligence systems also shows significant
promise in enhancing the collaborative intelligence of the systems, as shown in [13].
The literature suggests that the use of explainable systems, such as actionable explanations and feature
attribution, are critical in enhancing the trustworthiness of the systems, as shown in [2], [11], and [12]. The use
of highly accurate systems, such as the ones shown in [14] and [16], might also face difficulty in achieving the
trust and buy-in of the stakeholders, as the stakeholders might want explanations for the decisions made by the
systems on the crop yield and economic benefits.
As the literature suggests, the journey towards Agriculture 5.0, where human-centered, robust, and sustainable
smart farming systems can be achieved, as shown in [2] and [17], yet the systems are also highly fragmented,
and few studies show a unified approach towards the use of automation, HITL, and explainable systems.
As the discussion suggests, the future of precision agriculture might not only focus on the automation-driven
systems and the accuracy of the systems, but also on the use of collaborative intelligence systems, where human
expertise, explainability, and robustness are also considered.
Page 718
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
Table 2. Comparative analysis of fully automated artificial intelligence systems and Human-in-the-Loop
architecture in precision agriculture.
Fully Automated AI System
High accuracy in structured environments
Limited interpretability
Weak Adaptability to environmental variability
Reduced user trust
Static deployment pipeline
Figure 3. Partial Autonomy in HITL Systems [13]
Limitations And Future Work
This review recognizes several limitations, including the fact that it relies on a pre-selected list of 17
peerreviewed articles between 2020 and 2026. This may not represent the full scope of global research in this
rapidly advancing field of Human-in-the-Loop AI and precision agriculture. A second limitation of the review is
that, as a scoping review, it does not engage in any quantitative comparison of the accuracy of the models
presented, nor their effectiveness in real-world implementations.
A further limitation of the review is that the field of Human-in-the-Loop frameworks in agriculture remains in
its infancy, with many of the reviewed articles focusing on conceptual frameworks or small-scale
implementations of the technology. This means that long-term operational effectiveness, scalability, and
economic viability are still in their infancy.
Future research in the field of human-centered AI systems would benefit from developing frameworks that
integrate automation, human interaction, and explainable AI into a unifying framework that can be scaled up into
a cyberinfrastructure framework. Further emphasis would be placed on real-world field implementations, as well
as the development of adaptive feedback systems that can dynamically adjust the AI model based on expert
feedback. Further research into the socio-economic implications of human-centered AI systems would also
benefit the development of Agriculture 5.0 systems.
CONCLUSION
This scoping review aims to examine the role of Human-in-the-Loop Artificial Intelligence in supporting
precision agriculture through an examination of 17 studies on machine learning, IoT, collaborative robotics, and
explainability in AI systems. The study finds that although automation-based AI systems have shown significant
promise in enhancing crop monitoring, disease identification, and data-driven farm management, such systems
have shown limited potential in handling dynamic and ever-changing agricultural systems.
The integration of human expertise through HITL has shown promise in enhancing the robustness, reliability,
and adaptability of AI systems. Moreover, the role of explainability in AI systems has shown promise in
enhancing trust among stakeholders in AI-driven decision-making systems. Despite such advancements, the
integration of scalable infrastructure, adaptive human feedback, and explainability in AI systems is still in its
nascent stages in the literature.
The study concludes that the future of precision agriculture is not in automation systems, but rather in
collaborative intelligence systems that can address the challenges of sustainability, trust, and resiliency in smart
farming systems. The study finds that Agriculture 5.0 will be possible through the development of cohesive AI
systems that can address the challenges of sustainability, trust, and resiliency in smart farming systems.
Page 719
www.rsisinternational.org
INTERNATIONAL JOURNAL OF LATEST TECHNOLOGY IN ENGINEERING,
MANAGEMENT & APPLIED SCIENCE (IJLTEMAS)
ISSN 2278-2540 | DOI: 10.51583/IJLTEMAS | Volume XV, Issue II, February 2026
REFERENCES
1. Adve, S., et al. (2025). Advancing AI in agriculture through large-scale collaborative research.
Communications of the ACM.
2. Arafat, M. (2025). An explainable human-AI-in-the-loop framework for segmentation-aware
hyperspectral image classification. International Journal of Web and Open Systems.
3. Benos, L., et al. (2021). Machine learning in agriculture: A comprehensive updated review. Sensors,
21(11), 3758. https://doi.org/10.3390/s21113758
4. Benos, L., et al. (2025). Explainable AI-enhanced human activity recognition for human–robot
collaboration in agriculture. Applied Sciences, 15, 650. https://doi.org/10.3390/app15020650
5. Deka, G., et al. (2025). Enhancing precision agriculture through human-in-the-loop planning and
control. Proceedings of the IEEE Conference on Automation Science and Engineering (CASE 2024).
https://doi.org/10.1109/CASE59546.2024.10711319
6. Dolatabadian, A., et al. (2025). Image-based crop disease detection using machine learning. Plant
Pathology.
https://doi.org/10.1111/ppa.14006
7. Emon, M., et al. (2025). Integration of IoT, machine learning, and sensors for intelligent environmental
monitoring in agriculture. Journal of Computer Networks and Communications.
https://doi.org/10.1155/2025/ (Check article page for full DOI)
8. Holzinger, A., et al. (2025). Human-centered AI in smart farming: Toward Agriculture 5.0.
9. Mourtzinis, S., et al. (2025). A human-in-the-loop approach to applying large language models for farm
management insight.
10. Ngugi, L., et al. (2024). Machine learning and deep learning for crop disease diagnosis: A review.
Agronomy, 14. https://doi.org/10.3390/agronomy14123001
11. Razak, N. A., et al. (2024). Agriculture 5.0 and explainable AI for smart agriculture: A scoping review.
Emerging Science Journal.
12. Saranti, A., et al. (2025). Actionable explainable AI (AxAI): A practical example with aggregation
functions for adaptive classification. https://doi.org/10.3390/make4040047
13. Sreeram, V., & Nof, S. Y. (2021). Human-in-the-loop role in cyber-physical agricultural systems.
International Journal of Computers Communications & Control, 16(2).
https://doi.org/10.15837/ijccc.2021.2.4166
14. Subhan, F., et al. (2026). Crop disease detection using EfficientNetB0 deep learning architecture.
Discover Computing. https://doi.org/10.1007/s10791-025-09881-y
15. Sudha, M., & Loret, J. L. (2026). Machine learning-based precision agriculture techniques integrated
with IoT. Discover Environment. https://doi.org/10.1007/s44274-025-00305-8
16. Waltz, D., et al. (2025). Cyberinfrastructure for machine learning applications in agriculture. Frontiers
in Artificial Intelligence, 7.
https://doi.org/10.3389/frai.2025.1496066
17. Yakkala, K., et al. (2025). Deep learning-based crop health enhancement through early disease
prediction. Cogent Food & Agriculture, 11(1). https://doi.org/10.1080/23311932.2024.2423244