White-box Model Research Articles

Discovery of Highly Bioactive Peptides through Hierarchical Structural Information and Molecular Dynamics Simulations.

Peptide drugs play an essential role in modern therapeutics, but the computational design of these molecules is hindered by several challenges. Traditional methods like molecular docking and molecular dynamics (MD) simulation, as well as recent deep learning approaches, often face limitations related to computational resource demands, complex binding affinity assessments, extensive data requirements, and poor model interpretability. Here, we introduce PepHiRe, an innovative methodology that utilizes the hierarchical structural information in peptide sequences and employs a novel strategy called Ladderpath, rooted in algorithmic information theory, to rapidly generate and enhance the efficiency and clarity of novel peptide design. We applied PepHiRe to develop BH3-like peptide inhibitors targeting myeloid cell leukemia-1, a protein associated with various cancers. By analyzing just eight known bioactive BH3 peptide sequences, PepHiRe effectively derived a hierarchy of subsequences used to create new BH3-like peptides. These peptides underwent screening through MD simulations, leading to the selection of five candidates for synthesis and subsequent in vitro testing. Experimental results demonstrated that these five peptides possess high inhibitory activity, with IC50 values ranging from 28.13 ± 7.93 to 167.42 ± 22.15 nM. Our study explores a white-box model driven technique and a structured screening pipeline for identifying and generating novel peptides with potential bioactivity.

Improving rule-based classifiers by Bayes point aggregation

The widespread adoption of artificial intelligence systems with continuously higher capabilities is causing ethical concerns. The lack of transparency, particularly for state-of-the-art models such as deep neural networks, hinders the applicability of such black-box methods in many domains, like the medical or the financial ones, where model transparency is a mandatory requirement, and hence white-box models are largely preferred over potentially more accurate but opaque techniques.For this reason, in this paper, we focus on ruleset learning, arguably the most interpretable class of learning techniques. Specifically, we propose Bayes Point Rule Classifier, an ensemble methodology inspired by the Bayes Point Machine, to improve the performance and robustness of rule-based classifiers. In addition, to improve interpretability, we propose a technique to retain the most relevant rules based on their importance, thus increasing the transparency of the ensemble, making it easier to understand its decision-making process.We also propose FIND-RS, a greedy ruleset learning algorithm that, under mild conditions, guarantees to learn hypothesis with perfect accuracy on the training set while preserving a good generalization capability to unseen data points.We performed extensive experimentation showing that FIND-RS achieves state-of-the-art classification performance at the cost of a slight increase in the ruleset complexity w.r.t. the competitors. However, when paired with the Bayes Point Rule Classifier, FIND-RS outperforms all the considered baselines.

The cost of CO2 emissions abatement in a micro energy community in a Belgian context

This paper investigates and quantifies the benefits and the cost of CO2 emissions abatement of a Micro Energy Community (MEC) in a tiny residential cluster of three houses in a Belgian context. A simulation-based comparative analysis is performed of two individual and two MEC concepts, i.e. a reference scenario, an individual electrification scenario, an electricity sharing scenario, and an energy community scenario. A deterministic dynamic white-box modelling approach, including optimal control, is used, considering heat for space heating and domestic hot water, and electricity for electrical appliances. The energy system that achieves the greatest emission reduction at the lowest cost, has a collective heat pump and a photovoltaic installation sized based on the plug load demand. This system reduces the annual CO2 emissions by 11.48 tons at a cost of 179 EUR/ton CO2 considering 2021 dynamic retail electricity prices with a median of 251 EUR/MWh and a fixed retail gas price of 64 EUR/MWh. However, the results are strongly dependent on the gas-electricity price ratio. The highest value of the retail gas price in 2021 leads to a negative CO2 emissions abatement cost of 154 EUR/ton CO2, putting energy community concepts on the radar for a cost-effective energy transition.

Intelligent optimal control model of selection pressure for rapid culture of aerobic granular sludge based on machine learning and simulated annealing algorithm

Aerobic Granular Sludge (AGS) has advantages over Activated sludge (AS) but faces challenges with long granulation periods. In this study, a novel grey-box model is devised to optimize the cultivation of AGS to shorten the formation time. This model is based on an existing white-box model. The modeling process starts with the application of four sensitivity analysis methods to assess the 12 model metrics selected. Subsequently, 12 prediction models were constructed by combining the six Machine learning (ML) algorithms and integrated algorithms, with the best performance selected (R2 = 0.98). Finally, an AGS selection pressure planning model was designed in conjunction with a simulated annealing (SA) algorithm to guide AGS training. The results demonstrate that AGS formation could be achieved within four days under the model’s optimal control. Therefore, the establishment of this model provides a new technique for the cultivation of AGS.

Prediction of CO2 uptake in bio-waste based porous carbons using model agnostic explainable artificial intelligence

This study introduces comprehensive research on the prediction of the carbon dioxide (CO2) uptake from the biomass-waste derived-porous carbons (BWDPCs), by using scientometrics and model agnostic multi-layered explainable artificial intelligence (XAI) techniques. It aims to identify the main characteristics, and trends that are specific to this domain, and to establish, compare and analyse the four different black box machine learning (ML) models for CO2 uptake prediction. For this study, through model evaluation parameters, and scatter plots, statistical analysis supports the fact that the Extreme Gradient Boosting (XGBoost) model is found to be the best performing model for CO2 uptake prediction with low errors and high coefficient of correlation for both training (MSE: 0.157, RMSE: 0.397, MAE: 0.294, MAPE: 0.112, R2: 0.931) and testing phases (MSE: 0.345, RMSE: 0.588, MAE: 0.461, MAPE: 0.121, R2: 0.860). Now, with the best performing black box ML model as XGBoost model, it serves as the basis for the multi-layered XAI analysis. Using multi-layered XAI techniques to interpret the black box ML model and covert it to a white box model, it makes clearer insights into the significant key features that affect the CO2 uptake both at the global and local level. The study demonstrates that using multi-layered XAI analysis helps in improving the trust of the predictive model and provides a way forward for the application of white box models in CO2 uptake.

3 Hands -on 1: Applying system dynamics to develop “Flight Simulators” for sustainable animal production

Abstract Solving complex livestock production problems is a pressing issue for achieving long-term sustainability and profitability and often requires modeling techniques. The use of mathematical models is undoubtedly a daily practice in the livestock industry, especially nutrition, and has improved animal performance, productivity, and environmental sustainability while maintaining or reducing costs. Although many students and professionals use spreadsheets and existing empirical models for nutrition and management [NASEM (2016)], there is still a need to understand the complexity of livestock systems and the utility of flight simulator models. At the same time, more complex models (although robust) may fail to provide new insights for experienced nutritionists due to poor user-friendliness. A systems understanding goes beyond simply obtaining a desired output, such as optimizing a total mixed ration, but instead leads to identifying high-leverage solutions and gaining insight. Further, parameterizing and calibrating variables and equations and testing management scenarios is straightforward. However, developing causal feedback linkages (A to B and B to A) and identifying time delays is less intuitive and more challenging for novices. Model flight simulators grounded in fully documented, calibrated models provide a means to introduce practitioners to a methodology of insight generation because the user designs and runs the model scenarios for themselves, challenging their mental models. Such approaches are generally more impactful (compared with someone telling them) because they have gained insight into the system themselves, and, in the case of open source (white box models), they can explore equations and parameters. Therefore, understanding how to utilize dynamic models in scenario-based simulations is critical in training current and future modelers in animal science. This hands-on model training will cover the basics of System Dynamics modeling and allow participants to run real-time animal production simulations. Finally, participants will be “debriefed” to unpack “ah ha” moments that were unexpected. The debrief will include using models to develop accurate guidelines and recommendations for those who cannot use computer models and developing experimental designs to test hypotheses and “validate” model recommendations (proof in the pudding). Participants will gain knowledge of System Dynamics applications for animal production systems, experience using flight simulators, and their utility in teaching, informing, and guiding their livestock production or that of a client. The “flight simulator” will focus on production and nutrition with species-agnostic principles. Participants will also have a new tool to identify areas for improvement in model development (i.e., what is missing?), research questions, or industry needs. The need for modelers who can turn big data into insight, knowledge, and wisdom using a systems approach is becoming even more critical due to the increasing use of precision livestock farming. Thus, providing the livestock industry with trained systems modelers will help achieve current and future sustainability challenges.

Black Box Adversarial Reprogramming for Time Series Feature Classification in Ball Bearings’ Remaining Useful Life Classification

Standard ML relies on ample data, but limited availability poses challenges. Transfer learning offers a solution by leveraging pre-existing knowledge. Yet many methods require access to the model’s internal aspects, limiting applicability to white box models. To address this, Tsai, Chen and Ho introduced Black Box Adversarial Reprogramming for transfer learning with black box models. While tested primarily in image classification, this paper explores its potential in time series classification, particularly predictive maintenance. We develop an adversarial reprogramming concept tailored to black box time series classifiers. Our study focuses on predicting the Remaining Useful Life of rolling bearings. We construct a comprehensive ML pipeline, encompassing feature engineering and model fine-tuning, and compare results with traditional transfer learning. We investigate the impact of hyperparameters and training parameters on model performance, demonstrating the successful application of Black Box Adversarial Reprogramming to time series data. The method achieved a weighted F1-score of 0.77, although it exhibited significant stochastic fluctuations, with scores ranging from 0.3 to 0.77 due to randomness in gradient estimation.

FRRI: A novel algorithm for fuzzy-rough rule induction

Interpretability is the next frontier in machine learning research. In the search for white box models — as opposed to black box models, like random forests or neural networks — rule induction algorithms are a logical and promising option, since the rules can easily be understood by humans. Fuzzy and rough set theory have been successfully applied to this archetype, almost always separately. As both approaches offer different ways to deal with imprecise and uncertain information, often with the use of an indiscernibility relation, it is natural to combine them. The QuickRules [20] algorithm was a first attempt at using fuzzy rough set theory for rule induction. It is based on QuickReduct, a greedy algorithm for building decision superreducts. QuickRules already showed an improvement over other rule induction methods. However, to evaluate the full potential of a fuzzy rough rule induction algorithm, one needs to start from the foundations. Accordingly, the novel rule induction algorithm, Fuzzy Rough Rule Induction (FRRI), we introduce in this paper, uses an approach that has not yet been utilised in this setting. We provide background and explain the workings of our algorithm. Furthermore, we perform a computational experiment to evaluate the performance of our algorithm and compare it to other state-of-the-art rule induction approaches. We find that our algorithm is more accurate while creating small rulesets consisting of relatively short rules.

Learning metabolic dynamics from irregular observations by Bidirectional Time-Series State Transfer Network.

Modeling microbial metabolic dynamics is important for the rational optimization of both biosynthetic systems and industrial processes to facilitate green and efficient biomanufacturing. Classical approaches utilize explicit equation systems to represent metabolic networks, enabling the quantification of pathway fluxes to identify metabolic bottlenecks. However, these white-box models, despite their diverse applications, have limitations in simulating metabolic dynamics and are intrinsically inaccurate for industrial strains that lack information on network structures and kinetic parameters. On the other hand, black-box models do not rely on prior mechanistic knowledge of strains but are built upon observed time-series trajectories of biosynthetic systems in action. In practice, these observations are typically irregular, with discontinuously observed time points across multiple independent batches, each time point potentially containing missing measurements. Learning from such irregular data remains challenging for existing approaches. To address this issue, we present the Bidirectional Time-Series State Transfer Network (BTSTN) for modeling metabolic dynamics directly from irregular observations. Using evaluation data sets derived from both ideal dynamic systems and a real-world fermentation process, we demonstrate that BTSTN accurately reconstructs dynamic behaviors and predicts future trajectories. This approach exhibits enhanced robustness against missing measurements and noise, as compared to the state-of-the-art methods.IMPORTANCEIndustrial biosynthetic systems often involve strains with unclear genetic backgrounds, posing challenges in modeling their distinct metabolic dynamics. In such scenarios, white-box models, which commonly rely on inferred networks, are thereby of limited applicability and accuracy. In contrast, black-box models, such as statistical models and neural networks, are directly fitted or learned from observed time-series trajectories of biosynthetic systems in action. These methods typically assume regular observations without missing time points or measurements. If the observations are irregular, a pre-processing step becomes necessary to obtain a fully filled data set for subsequent model training, which, at the same time, inevitably introduces errors into the resulting models. BTSTN is a novel approach that natively learns from irregular observations. This distinctive feature makes it a unique addition to the current arsenal of technologies modeling metabolic dynamics.

Greybox-Modelle in der Zerspanung

Abstract Compared to uncoated tools, coated tools offer a longer tool life. However, it is difficult to predict the wear of coated tools using only empirical methods. The combination of deterministic whitebox models and data-driven blackbox models to greybox models can enable a prediction of the wear condition. Therefore, this article presents a concept for the application of greybox models for wear prediction.

Building retrofit optimization considering future climate and decision-making under various mindsets

Building retrofit is effective in reducing building energy use and improving comfort levels for existing buildings. However, conducting multi-objective optimization for individual buildings can be challenging due to the laborious computational cost of using white box models and the difficulty in visualizing and understanding the decision-making process. Additionally, the impact of climate change has not been fully considered for the post-retrofit lifecycle. This research proposes a pragmatic automated scheme that integrates a feature selection method based on marginal abatement cost analysis and variance-based sensitivity analysis, multi-objective optimization supported by non-dominated sorting differential evolution (NSDE) algorithm, tailor-made decision-making support under different mindsets, and tree-based retrospection scheme of decision-making pathways. The simulation engine used in this study is a low-order white box modeling tool developed by the research team. The proposed scheme was applied to two educational buildings with different thermal characteristics, and the results showed that a certain number of sampling sizes were needed to achieve reliable feature selection results. The hierarchical clustering based decision-making support scheme has demonstrated robustness in visualizing and supporting decision-making for Pareto front. Two retrofit mindsets - aggressive and balanced - were assumed in the decision-making process, and the proposed method produced distinct final solutions accoridng to the two mindsets. This framework can support informed decision-making, helping stakeholders implement sustainable practices and transition to a low-carbon built environment.

On the evaluation of surface tension of biodiesel

Over time, with the increase in population and the subsequent increase in energy consumption and also due to the non-renewability of fossil fuels, the study of alternative fuels has increased. One of these fuels is biodiesel, which is a suitable alternative to fossil fuels such as diesel and received much attention from researchers today. For this reason, measuring the physical properties of biodiesel is of great importance. Due to the high cost and time-consuming nature of laboratory methods, numerical methods are used to estimate material properties. The novelty of this research was the use of two white box models, including Group method of data handling (GMDH) and Gene expression programming (GEP), which work on the basis of artificial intelligence. By using these models, two simple mathematical equations with high accuracy were presented to predict the surface tension of biodiesel. These models can be used at different temperatures and molecular weights. To do modeling, 78 laboratory data available in the literature were gathered and the data were randomly divided into two groups, train and test, in a ratio of 80 and 20. The input parameters include mass fraction of fatty acid ethyl esters and temperature (T), and esters are divided into three groups according to their molecular weight: less than 200 (Mw1), between 200 and 300 (Mw2), and greater than 300 (Mw3). The statistical error parameters were calculated for the two models developed in this research and after comparing the results, it was found that the GMDH model estimates the surface tension of biodiesel with a higher accuracy. The average absolute relative error for GMDH and GEP models was reported as 0.97 and 1.89, respectively. Also, other statistical error parameters of GMDH such as RMSE, SD, and R2 for the GMDH model were obtained as 0.444, 0.000233, and 0.9233, respectively. Moreover, sensitivity analysis showed that temperature has the highest impact on the surface tension of biodiesel, which is also an inverse effect. Finally, suspicious laboratory and outlier data points were identified using the Leverage technique. According to this analysis, only five data points were identified as outliers and suspicious laboratory data.

Explainable and generalizable AI-driven multiscale informatics for dynamic system modelling

Ultra-precision machining requires system modelling that both satisfies explainability and conforms to data fidelity. Existing modelling approaches, whether based on data-driven methods in present artificial intelligence (AI) or on first-principle knowledge, fall short of these qualities in high-demanding industrial applications. Therefore, this paper develops an explainable and generalizable ‘grey-box’ AI informatics method for real-world dynamic system modelling. Such a grey-box model serves as a multiscale ‘world model’ by integrating the first principles of the system in a white-box architecture with data-fitting black boxes for varying hyperparameters of the white box. The physical principles serve as an explainable global meta-structure of the real-world system driven by physical knowledge, while the black boxes enhance local fitting accuracy driven by training data. The grey-box model thus encapsulates implicit variables and relationships that a standalone white-box model or black-box model fails to capture. Case study on an industrial cleanroom high-precision temperature regulation system verifies that the grey-box method outperforms existing modelling methods and is suitable for varying operating conditions.

Combining physical modeling and machine learning for micro-scale modeling of a fuel cell electrode

Abstract Microscale modeling plays a critical role in fuel cell development, offering deep insights into the microscale transport phenomena and electrochemical reactions. This level of detail is essential for optimizing the performance of a single fuel cell, enabling the precise design and improvement of materials and structures at the microscale and consequently enhancing the overall efficiency of a stack. Here, we show a comprehensive transition from white-box models, characterized by their reliance on physical laws, to black-box models exemplified by neural networks, which excel in pattern recognition from provided data without necessitating a clear understanding of the underlying processes. This spectrum encompasses the inherent challenges and merits of both methodologies. While white-box models are recognized for their reliability due to their foundation in mathematical equations that describe physical phenomena, they often require the integration of empirical parameters and are susceptible to experimental errors, much like their black-box counterparts. The core novelty in this study lies in the synergistic integration of these two paradigms, specifically tailored for enhancing the predictive accuracy in solid oxide fuel cell modeling. In this approach, the neural network is employed to replace different parts of the mathematical model, from refining empirical parameters in the electrochemical model to replacing the entire electrochemical model. The adjustment of parameters is conducted by an evolutionary strategy based on the outputs of the mathematical model. The results underscore the superiority of the gray box in achieving higher prediction accuracy and in minimizing the requisite data volume for network training. This presented approach not only bridges the gap between the deterministic clarity of white-box models and the data-driven insights of black-box models but also strategically distributes the computational load between them, thereby offering a promising solution to the prevalent challenges in solid oxide fuel cell modeling.

Stacking GA2M for inherently interpretable fraudulent reviewer identification by fusing target and non-target features

This paper proposes a novel approach called Stack-GA2M to identify fraudulent reviewers in an inherently interpretable manner by fusing both target and non-target features. Specifically, for local interpretability, we adopt GA2M (Standard Generalized Additive Model plus interactions) as the basic classifier to produce three subordinate models trained by using the target features and the non-target features as review textual features and reviewer behavioral features. For global interpretability, we adopt LR (Logistic Regression) as the meta classifier to stack the outputs of three subordinate models to identify the fraudulent reviewers. The white-box model of LR enables us to understand the global interpretability of the target features and the non-target features in identifying fraudulent reviewers. With GA2M, the local interpretability of each subordinate model is derived by using feature importance, spline shape functions for individual features, and heatmaps for interaction terms. Extensive experiments on Yelp dataset demonstrate that the proposed Stack-GA2M approach is superior to state-of-the-art techniques in identifying fraudulent reviewers and exhibits favorable inherent interpretability.

Pediatric cardiac surgery: machine learning models for postoperative complication prediction.

Managing children undergoing cardiac surgery with cardiopulmonary bypass (CPB) presents a significant challenge for anesthesiologists. Machine Learning (ML)-assisted tools have the potential to enhance the recognition of patients at risk of complications and predict potential issues, ultimately improving outcomes. We evaluated the prediction capacity of six models, ranging from logistic regression to support vector machine, using a dataset comprising 33 variables and 1364 subjects. TheArea Under the Curve (AUC)and theF1 scoreserved as the primary evaluation metrics. Our primary objectives were twofold: first, to develop aneffective prediction model, and second, to create auser-friendly comprehensive modelfor identifying high-risk patients. Thelogistic regression model demonstrated the highest effectiveness, achieving anAUC of 83.65%, and an F1 score of 0.7296, with balanced sensitivity and specificity of77.94%and76.47%, respectively. In comparison, thecomprehensive three-layer decision tree modelachieved anAUC of 72.84%, with sensitivity (79.41%) comparable to more complex models. Our machine learning-assisted tools provide an additional perspective and enhance the predictive capabilities of traditional scoring methods. These tools can assist anesthesiologists in making well-informed decisions. Furthermore, we have successfully demonstrated the feasibility of creating a practical white-box model. The next steps involve conducting clinical validation and multicenter cross-validation. NCT05537168.

Effect of prediction uncertainties on the performance of a white-box model predictive controller for district heating networks

The performance of a model predictive controller (MPC) depends on the accuracy of the controller model and predictions of system disturbances. This paper studies the effect of prediction uncertainties and/or an imperfect state observer on the performance of a white-box MPC applied to a small-scale district heating network by performing a co-simulation with a detailed emulator model. This effect is studied for both a non-linear program (NLP)-based MPC and a mixed-integer NLP (MINLP)-based MPC. The results show a limited impact of prediction uncertainties when using a perfect state observer, with electricity use and thermal discomfort changing by only ▪ to ▪ and ▪/day/building to 0.56 ▪/day/building. An MPC with perfect predictions and a simple state observer only impacts the MPC's trade-off between electricity use and thermal discomfort: the former increases by 2–▪ but the latter decreases by 0.01–▪/day/building. The combined effect of both simplifications has a significant impact with an increase in electricity use and thermal discomfort of 2–▪ and 0.35–▪/day/building, but MPC still outperforms a conventional rule-based controller. The MINLP-based MPC better predicts the actual system behaviour, but its performance is similar compared to the NLP-based MPC in terms of electrical energy use and thermal comfort.

Comparison of different control methods on the thermally activated building system (TABS) with large energy flexibility

Thermally Activated Building Systems (TABS) has great energy flexibility potential for the effective demand response, but the inherent large thermal inertia creates challenges for the control. Model Predictive Control (MPC) can improve the control efficiency while achieving certain objectives. However, the modeling mechanism of the TABS thermal behavior can lead to the difference in the controller performances. In this study, the white-box model, grey-box model, and black-box model based MPCs (referred to as W-MPC, G-MPC, and B-MPC) are developed and analyzed in depth for a case TABS with high heat flexibility potential. The performances of MPC strategies are investigated together with the baseline Rule Based Control (RBC) strategy under normal conditions as well as a variety of uncertainty events. In general, MPC strategies show better performance compared to the RBC. The prediction-based control can effectively maintain the indoor temperature within the constraints considering the uncertain disturbances and the gradual outdoor climate change. Moreover, MPC strategies help to increase the cost savings due to the effective load shifting by 30%–50%, and enhance the flexible energy utilization by 14%–29%. In terms of the comparison among the different MPC strategies, W-MPC shows better performance in the room temperature control, which reduces the violation of the indoor temperature constraint by 30% and 15% compared to G-MPC and B-MPC, respectively. G-MPC shows slight superiority concerning the economy and flexible energy usage. The normalized cost saving of G-MPC is approximately 3% and 8% lower compared to the W-MPC and B-MPC, respectively, while the utilization efficiencies of the flexible energy are approximately 6% and 12% higher.

A grey-box model with neural ordinary differential equations for the slow voltage dynamics of lithium-ion batteries: Application to single-cell experiments

Lithium-ion batteries exhibit a complex, nonlinear and dynamic voltage behaviour. Modelling their slow dynamics is a challenge due to the multiple potential causes involved. We present here a neural equivalent circuit model for lithium-ion batteries including slow voltage dynamics. The model uses an equivalent circuit with voltage source, series resistor, and diffusion element. The series resistance is parameterized using neural networks. The diffusion element is based on a discretized form of Fickian diffusion, parameterized using a neural network and learnable parameters. It is flexible to represent not only Warburg behaviour, but also resistor-capacitor-type dynamics. Mathematically, the resulting model is given by a differential–algebraic equation system combining ordinary and neural differential equations. Therefore, the model combines concepts of both physical theory (white-box model) and artificial intelligence (black-box model) to a combined framework (grey-box model). We apply this approach to a lithium iron phosphate based lithium-ion battery cell. The experimental voltage behaviour during constant-current cycles as well as the dynamics during pulse tests are well reproduced by the model. Only at very high and very low states of charge the simulation significantly deviates from experiments, which might result from insufficient training data in these regions.

Benchmarking Instance-Centric Counterfactual Algorithms for XAI: From White Box to Black Box

This study investigates the impact of machine learning models on the generation of counterfactual explanations by conducting a benchmark evaluation over three different types of models: a decision tree (fully transparent, interpretable, white-box model), a random forest (semi-interpretable, grey-box model), and a neural network (fully opaque, black-box model). We tested the counterfactual generation process using four algorithms (DiCE, WatcherCF, prototype, and GrowingSpheresCF) in the literature in 25 different datasets. Our findings indicate that: (1) Different machine learning models have little impact on the generation of counterfactual explanations; (2) Counterfactual algorithms based uniquely on proximity loss functions are not actionable and will not provide meaningful explanations; (3) One cannot have meaningful evaluation results without guaranteeing plausibility in the counterfactual generation. Algorithms that do not consider plausibility in their internal mechanisms will lead to biased and unreliable conclusions if evaluated with the current state-of-the-art metrics; (4) A counterfactual inspection analysis is strongly recommended to ensure a robust examination of counterfactual explanations and the potential identification of biases.