Black Box Research Articles

The Investigation of Artificial Intelligence in Cultural Relics Protection

It is well-known that the importance of cultural heritage in history, culture, art and other aspects is self-evident. However, due to various force majeure damage, the situation of cultural heritage protection is becoming more and more severe. This article will review and look forward to the application of Artificial Intelligence (AI) in cultural relics protection. The method section of this paper demonstrates various AI technologies such as Generative Adversarial Networks (GAN), deep learning, and reinforcement learning, and applies them to cultural relic digitization, damage detection, virtual restoration, and cultural heritage management and display. The next section outlines the specific tasks for which these AI techniques are applicable, demonstrating the potential of these innovations to improve accuracy and efficiency in work. In the subsequent discussion section, the current limitations and challenges of AI in this field are analyzed, such as data dependency and “black box” characteristics. At the same time, the future direction of AI is discussed, such as interpretable AI (XAI) and adaptive learning. Finally, this paper concludes that although AI has challenges, it will play an increasingly important role in the field of cultural heritage protection in the future.

The Promises and Challenges of AI-Facilitated Mental Health Care

Mental health is an increasingly critical issue that demands attention. With the rapid advancement of Artificial Intelligence (AI) across various domains, AI-facilitated mental health care is particularly promising and warrants further exploration in healthcare settings. This review provides insights into the current state of AI-facilitated mental health care, identifying both its potential benefits and the challenges that need to be addressed. Particularly during the COVID-19 pandemic, when access to traditional mental health services was limited due to social distancing rules and other restrictions, the potential of AI to facilitate mental health care effectively showed. Despite the promise of AI in mental health care, several challenges remain. The lack of transparency in many algorithms, often referred to as the “black box” issue, poses significant concerns. Additionally, the rapid development and continuous iteration of numerous mental health applications have resulted in a scarcity of comprehensive data to evaluate their effectiveness. The findings from this review can inform future research and practice in the mental health field, guiding the development of more effective and transparent AI-driven solutions to support mental health conditions.

Research on Explainability of Deep Neural Networks and Its Applications

In recent years, Artificial Intelligence (AI) models have surged in popularity and success, garnering much attention across many disciplines and applications. In particular, recent developments in Deep Neural Networks (DNN) have allowed them to match and even surpass human capabilities in many tasks. However, the complexity of deep neural networks causes them function like black boxes that are very difficult for humans to understand, thus making them untrustworthy and limiting their usefulness in high-risk applications. Explainable Artificial Intelligence aims to alleviate this issue by generating explanations for DNNs to allow humans to understand their reasoning and increase their transparency. This paper reviews research on neural network explainability, detailing its definition, necessity, classification, and evaluation. It categorizes current explanation methods for deep neural networks according to their number of inputs. Additionally, it introduces the principles and evaluation methods for common interpretability algorithms. Future research directions and applications of interpretable neural networks are summarized. The challenges facing interpretable neural networks are outlined, along with potential solutions to these challenges.

DEVELOPMENT OF AN E-COMMERCE PLATFORM USING EXTREME PROGRAMMING METHODOLOGY

<span lang="EN-US">The pervasiveness of information technology has reshaped human lifestyles, including shopping practices. Andalas Store, a prominent bookstore, has experienced a decrease in customer visits. Consequently, this research endeavours to establish an e-commerce platform for Toko Andalas utilizing Extreme Programming (XP) and the MERN Stack to broaden customer reach and expand market presence. The e-commerce platform is constructed leveraging the XP software development methodology and undergoes rigorous testing encompassing unit testing, black box testing, Lighthouse, and the System Usability Scale. The research outcomes validate the successful development and testing of the Toko Andalas e-commerce platform, highlighting its website performance in responsiveness, accessibility, adherence to best practices, and SEO optimization. The platform also encompasses comprehensive functionalities, including product search, shopping cart, and online payment. </span>

The Black Box of Single-Family Zoning Reform

America faces a staggering housing affordability crisis. The consequences of the crisis include reduced economic productivity, suppressed labor mobility, greater social inequity, and increased homelessness. A scholarly consensus blames excessive land-use regulation for these problems and advocates now call for deregulation. Several jurisdictions have recently elected to allow middle housing—small, multi-family developments—in areas previously zoned for single-family detached housing. The results of these reforms, however, have been disappointing. This Article examines why they have fallen short and how to do better. This Article identifies an analytical disconnect between the deregulatory consensus and actual single-family zoning reforms adopted to date. Given this disconnect and the limited success of single-family zoning reforms thus far, it is unclear why such reforms continue to be adopted. Moreover, these reforms cannot achieve their promised ends. Interviews of middle housing builders in reform jurisdictions reveal legal and structural barriers that prevent market participants from responding to the housing crisis with scale and urgency. This Article proposes several regulatory alternatives that offer a greater probability of success in achieving the consensus’s goals: (1) modestly expanding multi-family zones to allow redevelopment and densification of underutilized non-residential properties; (2) providing affordable-housing density bonuses; (3) streamlining land-use procedures and eliminating administrative delay to reduce uncertainty in permitting processes; and (4) reevaluating financing mechanisms for new housing development. Taken together, these policies are necessary complements to single-family zoning reforms for addressing undersupply of market- and below-market-rate housing.

Phase I results of both IV and oral administration for CG-0255: a novel fast-acting antiplatelet drug with a new activation pathway

Abstract Clopidogrel is an extremely widely used antiplatelet drug, but with serious shortcomings. It is a prodrug that depends critically on liver CYP oxidation enzyme CYP2C19 for activation to its active metabolite H4. However, large fractions of world population carry loss of function CYP2C19 alleles, ranging from a low ~21% in Latino to a strikingly high ~59% in east Asian population. Known as clopidogrel resistance, US FDA added a black box warning to clopidogrel label. Despite this issue, routine CYP2C19 screening for clopidogrel patients is not performed, putting many patients at risk. Clopidogrel is also slow to take effect and not amenable for IV formulation, limiting its clinical use in emergency. To overcome these issues, we have invented a novel antiplatelet drug CG-0255, which is the world’s first thioester prodrug with a new activation pathway. CG-0255 has been designed to require only abundantly present and widely distributed carboxylesterases for metabolizing to its active metabolite, which is identical to clopidogrel active metabolite H4 generated through CYP metabolism. Thus, CG-0255 completely bypasses the CYP metabolism and overcomes clopidogrel resistance. In addition, CG-0255 has a much faster onset of action and high solubility in water and has been formulated for both IV and oral administration. Here we report the results of an open label phase I study of CG-0255 oral tablets to assess safety, tolerability, pharmacodynamics (PK), and pharmacokinetics (PD: inhibition of platelet aggregation, IPA) in healthy volunteers. It was a single dose escalation study, 4 cohorts with 6 participants each given CG-0255 oral tablets at doses ranging from 2 to 10 mg. Clopidogrel was the positive control and given at 300 mg. Blood samples for PK/PD analysis were collected pre-dosing and at various time points post dosing. VerifyNow was used for IPA measurement. CG-0255 is generally safe and well tolerated. PK analysis shows fast and consistent conversion of CG-0255 to active metabolite H4. PD results indicate that CG-0255 IPA is dose-dependent, easily matching or surpassing that of clopidogrel 300 mg for all doses. At the highest 10 mg dose, CG-0255 IPA reaches over 90% in less than 30 minutes, well above that of clopidogrel. In addition, CG-0255 IPA shows minimum individual variations, and the antiplatelet effect is long lasting due to its irreversible binding to the P2Y12 receptor. These results are consistent with our earlier reported Phase I results for CG-0255 IV injection. In conclusion, CG-0255 is a novel, fast-acting and long-lasting antiplatelet drug. It shares identical metabolite H4 with clopidogrel but relies only on carboxylesterases for activation, overcoming the resistance and other issues associated CYP-based oxidation metabolism. As the only antiplatelet drug available for both IV and oral administrations, CG-0255 fills unmet medical needs and will greatly benefit patients and medical community.

Marketing or health promotion? A citizen science study of social media influencers’ health content

Abstract Background Social media influencers (SMI) became the dominant source of health information among adolescents, exerting influence through appealing content and para-social relationships. Personalized social media feeds of adolescents paired with latent platform algorithms result in myriads of individual information environments, which remain closed to researchers. Citizen science offers a solution to peek inside this black box and identify: What health content is disseminated by social media influencers and consumed by adolescents? Methods The citizen science study adopted a contributive approach, where adolescents collected and categorized SMI profiles and posts from their own feeds with a pre-tested survey tool which was available on Google Tabs. Data collection was conducted on Instagram, TikTok, and YouTube between April and May 2023 and subsequently analyzed by the research team. Results The study involved 19 school classes, containing 326 students, aged 12 to 19 years. They documented 255 SMI profiles and 2147 posts, primarily on Instagram and TikTok. Key topics included fitness, beauty ideals, and nutrition, while mental and sexual health received less attention. Although 41.5% of the posts contained advertising, the citizen scientists classified 63.5% as health-promoting. Topics relating to dietary supplements (84.1%), addictive substances (55.6%), nutrition (53.8%) and sport (47.2%) were particularly ad-heavy. Despite mental and sexual health having the lowest advertising rates at 35.4% and 37.7% respectively, this still accounts for over one third of the documented posts. Conclusions SMI present a distorted image of health that emphasizes nutrition, fitness and beauty ideals, while topics such as mental and sexual health are neglected. The positive reception of posts containing ads indicates a lack of critical reflection. Policies for advertising practices and balanced health education, including media literacy, are necessary to address these trends. Key messages • Social media influencers propagate a distorted health image, defined by nutrition, fitness, and beauty ideals. • Adolescents lack the necessary skills to critically reflect on ad littered social media influencer health content.

Portfolio location-related decisions: do proximities matter?

PurposeThis paper aims to examine the relationships between how family owners perceive different aspects of proximity to their territory of origin and the location-related decisions of family business portfolios.Design/methodology/approachBy using the data collected from 112 family business portfolios (FBP), the authors carry out an empirical analysis to test the proposed hypotheses. The results are robust across a battery of robustness analyses performed by the authors.FindingsThe analysis reveals a heterogeneous relationship between the different dimensions of proximity and the maintenance and location of portfolio business units. Specifically, we reveal that entrepreneurial families’ social and institutional proximities to their founding territory have a positive influence on their decisions related to investing and maintaining their business portfolio.Originality/valueThese findings open the black box of how business decisions are affected by how business owners perceive their contexts, providing new insights into how corporate decisions in understudied business organisations, business groups, are shaped by the ownership’s proximity to territories.

NutriCare: Aplikasi Pemantau Gizi Ibu Hamil untuk Menghindari Stunting dengan Memenuhi Kebutuhan Kalori dan Zat Besi

Stunting in children under five, due to chronic malnutrition since in the womb, has an impact onchildren's health and cognitive development as well as the potential for long-term economic losses.In Indonesia, the prevalence of stunting will reach 21.6% in 2022. Iron and calorie deficiency inpregnant women is the main factor causing stunting, with 48.6% of pregnant women experiencinganemia. Many pregnant women lack easily accessible nutritional information. The Hilda AlnairaClinic in Koja, North Jakarta, faces challenges in providing manual nutritional recommendationsthat are inefficient and lack personalization. This research aims to develop the "NutriCare" mobileapplication to help pregnant women monitor calorie and iron intake. This Android-basedapplication was developed using a prototypemethod, easy-to-use interface, relevant features, testedusing the black box method, and optimized before full implementation. This application helpsincrease pregnant women's understanding of nutrition and supports maternal health and fetaldevelopment in efforts to prevent stunting.

Algorithmic Bias in BERT for Response Accuracy Prediction: A Case Study for Investigating Population Validity

AbstractPretrained large language models (LLMs) have gained popularity in recent years due to their high performance in various educational tasks such as learner modeling, automated scoring, automatic item generation, and prediction. Nevertheless, LLMs are black box approaches where models are less interpretable, and they may carry human biases and prejudices because historical human data have been used for pretraining these large‐scale models. For these reasons, the prediction tasks based on LLMs require scrutiny to ensure that the prediction models are fair and unbiased. In this study, we used BERT—a pretrained encoder‐only LLM for predicting response accuracy using action sequences extracted from the 2012 PIAAC assessment. We selected three countries (i.e., Finland, Slovakia, and the United States) representing different performance levels in the overall PIAAC assessment. We found promising results for predicting response accuracy using the fine‐tuned BERT model. Additionally, we examined algorithmic bias in the prediction models trained with different countries. We found differences in model performance, suggesting that some trained models are not free from bias, and thus the models are less generalizable across countries. Our results highlighted the importance of investigating algorithmic fairness in prediction models utilizing algorithmic systems to ensure models are bias‐free.

The Underlying Cognitive Processes of Thin Slices Judgments on Teaching Quality

Thin slices ratings (i.e., ratings based on first impressions) have yielded intriguingly accurate results in various domains. Among other, researcher have applied the thin slices technique to assess instructional quality, showing that teacher-student interactions can be reliably inferred by just very short snippets of classroom instruction. The accuracy of thin slices ratings is often explained by dual process theories of social cognition, whereby System 1 refers to an intuitive and fast way of processing, while System 2 denotes a more reflective and analytical way of processing. System 1 is considered the cognitive foundation of thin slices ratings. The central aim of the present study was to understand the underlying cognitive processes shaping the impression formation of thin slices raters of teaching quality. Therefore, an unconventional and innovative research design was required to gain insights into the cognitive “black box” of thin slices raters by examining their verbal data. In an exploratory mixed method research design, we set up Cognitive Laboratories with two different rating situations. In a thin slices rating situation, participants rated instructional quality based on short classroom videos (30 seconds). Participants in a long-video rating situation rated instructional quality based on longer classroom videos (10 minutes). We collected, coded and statistically analyzed participants’ verbal reports regarding their rating processes. The findings suggest that thin slices ratings evolve primarily based on typical processes of System 1 and not on those of System 2. For instance, thin slices ratings are associative and tend to be rather negative than positive. Moreover, an initially formed impression tends to remain stable and is resistant to alteration. Ratings of instructional quality based on longer videos rely on both cognitive systems, with System 2 possibly modifying an initial judgment. Thus, our study does not only explain the cognitive processes under-lying the thin slices ratings, but additionally provides valuable insights into the processes occurring in conventional rating settings.

A Critical Assessment of the Process and Logic Behind Fish Production in Marine Recirculating Aquaculture Systems

A recirculating aquaculture system (RAS) represents a forward-looking form of aquaculture. A RAS consists of fish tanks and water treatment processes in a closed loop to sustain the environmental conditions for fish production. However, the rapid industrialization of the technology is fraught with transfer problems. This review justifies a RAS process chain based on fish biology. The underlying concept has been evaluated by the author in experimental and commercial RAS projects. The core idea is that the fish must be considered as a technical subcomponent in a RAS, determining the technology. Fish, when considered as small biological machines, are still a black box in many ways. However, their basic biology and physiology provide all the knowledge to implement them in a technical setting. The information required to understand this concept is presented and discussed based on current scientific knowledge. The conclusion is that the technology is available but needs to be rigorously implemented. If this were carried out, fish production in RASs would be ecologically sustainable, which is already claimed for RASs but is not always the reality in commercial applications.

PERANCANGAN SISTEM ENTERPRISE RESOURCE PLANNING (ERP) TERHADAP PENGADAAN BAHAN BAKU, PENJUALAN, DAN PRODUKSI PADA CV. JAYA LOGAM MENGGUNAKAN SOFTWARE ODOO

CV. Jaya Logam faced various operational challenges caused by ineffective sales and raw material inventory recording systems, leading to production delays, errors in information delivery, and difficulties in monitoring and decision-making. This research was conducted to address these issues through system needs analysis, design, and testing of an Enterprise Resource Planning (ERP) system using Odoo software. The methodological process involved a detailed analysis of system requirements, system design using Unified Modelling Language (UML) diagrams to visualize the system architecture, and system testing using the Black Box Testing method to ensure the reliability and effectiveness of the designed system. The testing results showed that the developed ERP system successfully met all 12 test class parameters, indicating that the system is valid and can be implemented effectively. The implementation of this ERP system at CV. Jaya Logam resulted in significant improvements in operational efficiency and data accuracy, better business process integration, and the provision of real-time information that supports faster and more accurate decision-making processes.

Request Deadline Split and Interference‐Aware Request Migration in Edge Cloud

ABSTRACTEdge computing extends computing resources from the data center to the edge of the network to better handle latency‐sensitive tasks. However, with the rise of the Internet of Things, edge devices with limited processing capabilities face difficulties in executing requests with fluctuating request peaks. In order to meet the deadline constraints of latency‐sensitive tasks, a feasible solution is to offload some latency‐sensitive tasks to other nearby edge devices. This article studies the problem of request migration in edge computing systems and minimizes the request deadline violation rate based on actual online arrival patterns, performance interference phenomena, and deadline constraints. Since a request contains multiple services and request migration will lead to changes in server resource competition pressure, we split the problem into three sub‐problems, dividing the request deadline to determine the maximum response time of the service, determining the performance of the service under different resource pressures and the request migration strategies. To this end, we propose two deadline splitting methods, a performance interference model under multi‐resource pressure, and two heuristic request migration strategies. Since this article considers online edge scenarios, the number and type of requests are black boxes. We conduct simulation experiments and find that our method has only one‐third the number of request violations of other methods.

Data-driven discovery of chemotactic migration of bacteria via coordinate-invariant machine learning

BackgroundE. coli chemotactic motion in the presence of a chemonutrient field can be studied using wet laboratory experiments or macroscale-level partial differential equations (PDEs) (among others). Bridging experimental measurements and chemotactic Partial Differential Equations requires knowledge of the evolution of all underlying fields, initial and boundary conditions, and often necessitates strong assumptions. In this work, we propose machine learning approaches, along with ideas from the Whitney and Takens embedding theorems, to circumvent these challenges.ResultsMachine learning approaches for identifying underlying PDEs were (a) validated through the use of simulation data from established continuum models and (b) used to infer chemotactic PDEs from experimental data. Such data-driven models were surrogates either for the entire chemotactic PDE right-hand-side (black box models), or, in a more targeted fashion, just for the chemotactic term (gray box models). Furthermore, it was demonstrated that a short history of bacterial density may compensate for the missing measurements of the field of chemonutrient concentration. In fact, given reasonable conditions, such a short history of bacterial density measurements could even be used to infer chemonutrient concentration.ConclusionData-driven PDEs are an important modeling tool when studying Chemotaxis at the macroscale, as they can learn bacterial motility from various data sources, fidelities (here, computational models, experiments) or coordinate systems. The resulting data-driven PDEs can then be simulated to reproduce/predict computational or experimental bacterial density profile data independent of the coordinate system, approximate meaningful parameters or functional terms, and even possibly estimate the underlying (unmeasured) chemonutrient field evolution.

LLpowershap: logistic loss-based automated Shapley values feature selection method

BackgroundShapley values have been used extensively in machine learning, not only to explain black box machine learning models, but among other tasks, also to conduct model debugging, sensitivity and fairness analyses and to select important features for robust modelling and for further follow-up analyses. Shapley values satisfy certain axioms that promote fairness in distributing contributions of features toward prediction or reducing error, after accounting for non-linear relationships and interactions when complex machine learning models are employed. Recently, feature selection methods using predictive Shapley values and p-values have been introduced, including powershap.MethodsWe present a novel feature selection method, LLpowershap, that takes forward these recent advances by employing loss-based Shapley values to identify informative features with minimal noise among the selected sets of features. We also enhance the calculation of p-values and power to identify informative features and to estimate number of iterations of model development and testing.ResultsOur simulation results show that LLpowershap not only identifies higher number of informative features but outputs fewer noise features compared to other state-of-the-art feature selection methods. Benchmarking results on four real-world datasets demonstrate higher or comparable predictive performance of LLpowershap compared to other Shapley based wrapper methods, or filter methods. LLpowershap is also ranked the best in mean ranking among the seven feature selection methods tested on the benchmark datasets.ConclusionOur results demonstrate that LLpowershap is a viable wrapper feature selection method that can be used for feature selection in large biomedical datasets and other settings.

Selection of aerodynamic characteristics and engine parameters of a maneuverable aircraft under epistemic uncertainty

At the preliminary stage of aircraft design, it is usually required to solve the problem of insufficient initial data for applying traditional algoristic-type mathematical programming models. In many cases, numerous input parameters cannot be accurately specified at the time of making design decisions. If the inaccuracy of parameters is not taken into account, the actual values of target functions may differ significantly from the calculated ones when solving optimization problems. In this regard, the issue of current interest is the development of algorithms to improve the reliability of design decisions under conditions of epistemological uncertainty when experts engage in the formation of initial data. The paper considers the problem of selecting aerodynamic characteristics and engine parameters of maneuverable aircraft under conditions of uncertainty associated with inaccuracy of expert data. The applied problem under consideration is further complicated by the necessity to include “black box” models in the algorithms being developed. The paper proposes algorithms that apply the uncertainty theory together with “black box” models that implement optimization calculation techniques derived from previous engineering practice of aircraft design. Using these algorithms, experts are able to set uncertain parameters whereby the lack of data is factored in using uncertainty distribution functions. In cases of monotonicity of uncertain parameters in target functions, application of the uncertainty theory provides a significant reduction in computational costs compared to the method of statistical modeling in optimization calculations. The paper presents the results of computational studies of the developed algorithms. A genetic algorithm (mathematical optimization solver) is used to optimize the search for design solutions. Pareto frontiers are obtained for different confidence levels enabling to make design decisions including values of aerodynamic characteristics and engine parameters of maneuverable aircraft.

De- and recoding algorithmic systems: The case of fact checkers and fact checked users

With the recent development of debunking on social media as a dominant agenda, fact checkers have increasingly used machine learning (ML) to identify, verify and correct factual claims, as ML promises the scaling of fact checking practices. However, it also places a new actor in between the fact checkers and the fact checked users. In this paper, we conducted a contrasted analysis of how fact checkers and fact checked users understand, evaluate and act towards the algorithmic system and the data flows in Meta’s Third-Party Fact-Checking Program: We did ethnographic fieldwork in the fact checking newsroom and interviewed and did protocol analysis with the fact checked. For both professional users and end users, the algorithmic system is experienced as a black box in which they have limited insight, and their sense-making practices happen based on the data and metrics that are made visible to them. In the paper, we draw on and expanded theory on decoding algorithms by not only exploring how the two user groups engage in decoding the algorithmic system, but also actively engage in forms of recoding by attempting to adapt or modify the algorithmic system to better fit within their cultural and social context, which is characterised by both varying epistemic cultures and societal positions. While the fact checkers from their hegemonic (sometimes negotiated) position understand the program as a (sometimes stupid) tool and primarily engage in passive acts of recoding, fact checked users, from their oppositional position, understand the program as an unpredictable censoring machine and engage primarily in more active acts of recoding. Based on the analysis, we end the paper with a discussion in which we argued for understanding data reflexivity as highly relational and processual.

ASIP Performance Enhancement by Hazard Control through Scoreboard

The application-specific instruction set processor (ASIP) has been gradually accepted in AI, communication, media, game and industry control. The digital signal processor (DSP) is a typical ASIP, whose benefits include high performance in specific domains, low power consumption, high flexibility and low silicon consumption. One of the challenges for DSP design is to handle problems induced by datapath acceleration. The datapath acceleration (instruction fusion, black box instructions) induces control complexities. To most efficiently utilize hardware, control challenges can be summarized as RAW (Read-After-Write) handling, hardware hazard handling, and WAW (Write-After-Write) handling. Both an advanced compiler and hardware hazard handler can be used as solutions. In this paper, we introduced both solutions and exposed the benefits from the hardware solution. The benefits include utilizing low silicon to achieve higher performance and program memory reduction on chip. In summary, our solution only uses 0.91% extra silicon area yet achieves 32.75% performance improvement. So, the overall performance-to-cost ratio could be evaluated as 32%.

Exploring the black box of human reproduction: endometrial organoids and assembloids - generation, implantation modeling, and future clinical perspectives.

One of the critical processes in human reproduction that is still poorly understood is implantation. The implantation of an early human embryo is considered a significant limitation of successful pregnancy. Therefore, researchers are trying to develop an ideal model of endometrium in vitro that can mimic the endometrial micro-environment in vivo as much as possible. The ultimate goal of endometrial modeling is to study the molecular interactions at the embryo-maternal interface and to use this model as an in vitro diagnostic tool for infertility. Significant progress has been made over the years in generating such models. The first experiments of endometrial modeling involved animal models, which are undoubtedly valuable, but at the same time, their dissimilarities with human tissue represent a significant obstacle to further research. This fact led researchers to develop basic monolayer coculture systems using uterine cells obtained from biopsies and, later on, complex and multilayer coculture models. With successful tissue engineering methods and various cultivation systems, it is possible to form endometrial two-dimensional (2D) models to three-dimensional (3D) organoids and novel assembloids that can recapitulate many aspects of endometrial tissue architecture and cell composition. These organoids have already helped to provide new insight into the embryo-endometrium interplay. The main aim of this paper is a comprehensive review of past and current approaches to endometrial model generation, their feasibility, and potential clinical application for infertility treatment.