Deceptive Strategies Research Articles

Optimizing Financial Fraud Detection: Understandings from Variable Selection with Neutrosophic Vague Soft Set

Neutrosophy is the neutralities study and prolongs the discussion of the truth of opinions. Neutrosophic logic might be used in all sectors, to provide the solution for the indeterminate challenges. Some real-time data experience issues like inconsistency, incompleteness, and indeterminacy. A fuzzy set (FS) offers an uncertain solution, and an intuitionistic fuzzy set (IFS) processes partial data, but both fail to handle uncertain data. Financial fraud, believed as a deceptive strategy to gain financial assistance, has recently become a common threat in organizations and companies. Traditional methods namely manual inspections and verifications are costly, time-consuming, and imprecise to identify such fraudulent actions. With the development of artificial intelligence (AI), machine learning (ML)-based algorithms are applied logically to identify fraud transactions by investigating a larger amount of financial data. Therefore, the study offers an Optimizing Financial Fraud Detection using Bayesian Optimization and Variable Selection with Neutrosophic Vague Soft Set (OFFDBO-VSNVS) Algorithm. The OFFDBO-VSNVS model presents an optimized framework for fraud detection by integrating advanced variable selection techniques and classification models. Initially, the OFFDBO-VSNVS technique applies the Z-score data normalization technique to transform input data into a compatible layout. Next, the grey wolf optimizer (GWO)--based feature selection to effectively reduce dimensionality and highlight the most relevant features. For the classification and detection of financial fraud, the neutrosophic vague soft set (NVS) model can be employed. Eventually, the Bayesian optimization (BO) model adjusts the hyperparameter values of the NVS algorithm optimally and outcomes in greater classification performance. The stimulated outcome study of the OFFDBO-VSNVS model occurs and the outcomes are examined in terms of changing features. The experimental study represented the superiority of the OFFDBO-VSNVS method across the existing state-of-the-art methods

Produktives Täuschen. Dissimulatio artis zwischen Rhetorik und Philosophie

Abstract In our paper, we discuss the role of deception (apáte) in the conflict between rhetoric and philosophy. While Parmenides and Plato seek to escape the possibility of being deceived in general, their counterparts, the sophists and early theorists of rhetoric, emphasize that this is neither possible nor desirable. In contrast to their philosophical contemporaries, they do not assume a dualism of truth and deception, but rather a gradualist spectrum that contains many shades of gray. Furthermore, they emphasize a certain productivity of particular deceptions that can help us increase our imagination and prefigure possibilities of life and praxis that would not be possible without being deceived. Moreover, we ask whether and how the potential of classical rhetoric for theorizing deception can be used for a critique of deception that is aware of its own proneness to illusions. The guiding hypothesis is that rhetoric, as a technique of verbal deception, can also provide an important key for criticizing deceptive strategies. This applies in particular to the concept of dissimulatio artis, which goes back to Aristotle’s Rhetoric and had been developed further by Quintilian, Ovid, and the anonymous author of the treatise Perí hýpsous. This concept makes it possible to describe simulations of unsimulatedness in philosophy as well as in politics.

Pollination‐related plant traits under environmental changes: Seasonal and daily mismatches produce temporal constraints

Abstract Pollination is a key tenet of ecosystem sustainability and food security, but it is threatened by climate change. While many studies investigated the response of plant‐pollination traits to temperature, few attempted multifactorial and integrative approaches with multiple floral traits. We determined which plant‐pollination traits were disrupted by environmental changes and the consequence for plant fitness. We focused on the thermogenic plant, Arum italicum, which relies on floral scent and heat production to attract its pollinators with a deceptive strategy. In a field mesocosm, we measured the response of floral phenology, thermogenesis traits, scent and plant fitness to experimental changes in shade level, water supply and soil temperature. In the laboratory, we exposed plants to extreme heat to determine the thermoregulation ability at high temperatures. Plant fitness decreased dramatically in response to soil warming and the level of shade. While both factors caused a lower number of inflorescences, soil warming was mostly responsible for the crash in number of fruits. The plant traits related to thermogenesis and scent were relatively resilient to the environmental changes in the mesocosm and were unlikely to produce the observed decrease in plant fitness. First, environmental changes only weakly modified the scent composition. Second, the appendix regulated its maximal temperature almost perfectly, while stamen regulated imperfectly. In the laboratory experiment, the appendix maintained a thermal gradient supposedly to improve scent emission while stamen became cooler than ambient air, possibly preventing the pollinators from overheating within the floral chamber. By contrast, the phenology of flowering was drastically modified. Flowers appeared earlier in the season in response to soil warming (by 39 days) or absence of shade (by 19 days). At the daily scale, flowering hour did not change across the season resulting in plants flowering before sunset late in the season and before the pollinator activity window (mostly nocturnal). The strong shift observed in the flowering period may induce a temporal mismatch with this pollinator both early and late in the season when the insects may not be active so early in the spring and during daytime, respectively, thereby altering plant fitness. Read the free Plain Language Summary for this article on the Journal blog.

The deceptive pollination of Anacamptis morio: generalised food deception or an initial step towards adaptive mimicry?

ABSTRACT Background To attract their pollinators non-rewarding orchids rely on various deception strategies that range from generalised food deception to floral mimicry of a specific model (Batesian mimicry). Aims We evaluated whether the deceptive orchid Anacamptis morio may represent an initial step in the evolutionary trajectory from generalised food deception to Batesian mimicry by resembling the rewarding Polygala nicaeensis, a species with purple flowers very similar to those of A. morio. Methods We assessed the pollination success of A. morio in relation to co-flowering rewarding species. Then, we estimated male and female reproductive success of A. morio depending on the presence of and similarity to the putative model P. nicaeensis. Results The pollination success of A. morio covaried with that of P. nicaeensis and its pollination efficiency was higher when the two species co-occurred. However, we did not find supporting evidence that A. morio mimicked its potential rewarding model. We also found that the orchids with stronger colour reflectance exhibited greater pollination success, according to a scenario of ecological facilitation rather than adaptive mimicry. Conclusion Co-flowering rewarding species with a colour similar to the orchid may enhance the specific pollinator abundance and condition their foraging preference, hence producing a positive effect on the reproductive success of A. morio.

Unveiling Greenwashing Strategies: A Comprehensive Analysis of Impacts on Consumer Trust and Environmental Sustainability

This document explores the phenomenon of greenwashing, a deceptive practice where companies project an environmentally responsible image without substantial environmental benefits that involve strategies like misleading labels, vague claims, and irrelevant or false assertions to appeal to eco-conscious consumers. The research aims to unravel the deceptive strategies employed by companies across various industries, understand the implications of these practices on consumers, investors, and the environment, and assess the effectiveness of current regulatory frameworks in mitigating these activities. This qualitative study employs a thematic analysis of the literature to categorize greenwashing strategies and assess their impacts on consumer trust, environmental sustainability, and corporate accountability. The findings reveal that greenwashing undermines consumer trust, genuine sustainability efforts, and informed decision-making. The study seeks to offer actionable insights for policymakers, businesses, and consumers to address and mitigate the effects of greenwashing. It emphasizes the need for stricter regulations and consumer education to counteract greenwashing. The study also highlights the role of technological advancements like blockchain and IoT in enhancing transparency. The study calls for a multifaceted strategy to address greenwashing, focusing on transparency, accountability, and genuine environmental responsibility.

Is India on a path to reduce the tobacco industry's influence in tobacco control? Insights from the Global Tobacco Industry Interference Index (2019-2023).

The tobacco industry (TI), driven by profit motives, consistently conceals health risks through deceptive strategies, notably in countries like India. These tactics create vulnerabilities that hinder effective tobacco control measures and enable the TI to exploit legal gaps. Understanding these TI strategies is essential for policymakers to take appropriate preventive and corrective measures in order to limit tobacco industry interference (TII) in policy-making. The study aims at understanding the trend of TII in India between 2019 and 2023. The secondary data from the Global Tobacco Industry Interference report, consisting of seven major domains of the TII index, viz. policy participation, corporate social responsibility (CSR) activities, industry benefits, unnecessary interaction, transparency, conflict of interest, and preventive measures, were retrieved. A composite score was obtained after adding scores of different domains, for each year. The findings of the study demonstrated an initial improvement in India's implementation of WHO FCTC Article 5.3, as evidenced by a decreasing score between 2019 and 2021. However, this trend halted in 2023, with data showing a slight increase in the score. When compared with other Asian countries, India shows marginal improvement in score than Cambodia, China, Indonesia, Lao PDR, Malaysia, Myanmar, Nepal, Philippines, Thailand, and South Korea. Some of the countries in the region, including India, Pakistan, Bangladesh, Sri Lanka, Myanmar, Brunei, China, and Vietnam experienced a decline in TII. There has been a rise in CSR activities, forms of unnecessary interactions of TII with policymakers, and participation in policy development; however, improvements are observed in providing benefits to the TI, conflict of interest, and preventive measures. In order to fortify the regulatory framework, it is imperative to create awareness among stakeholders on conflict of interest, denormalize corporate social responsibility (CSR) initiatives by the TI, provision of a watchdog for TII in the country and "whole of government" approach in implementation of FCTC Article 5.3.

Eco-Frauds: The Ethics and Impact of Corporate Greenwashing

The evolving dynamics of the marketplace, coupled with concerns regarding the finite capacity to meet increasing demands, have led to the emergence of new phenomena and practices. These developments, while heralding significant changes in the perception and selection of products and services, also elicit substantial concerns. Greenwashing is defined as the strategic practice by which corporations create a misleading impression of their environmental initiatives. This paper examines the theoretical foundations and multifaceted nature of greenwashing, identifying key deceptive strategies such as hidden trade-offs, lack of verifiable evidence, vagueness, irrelevance, and false labels. It explores the motivations behind greenwashing, which often stem from the pressure to appear environmentally responsible without incurring the costs associated with genuine sustainability. The analysis underscores the broader implications of greenwashing, including its impact on consumer trust, corporate governance, and environmental policy. Additionally, the paper addresses the emerging issue of "machinewashing," wherein companies make exaggerated claims about the ethical aspects of their AI technologies. Furthermore, the concept of environmental vices is explored, illustrating how greenwashing perpetuates unethical behaviors and undermines authentic sustainability efforts. To address these practices, the study advocates for a comprehensive approach that includes stricter regulations, increased transparency, and greater engagement from civil society. By promoting genuine sustainability and holding corporations accountable, it is possible to restore consumer trust and support authentic environmental initiatives, thereby fostering a more sustainable and ethical commerce.

A benchmark dataset for community deception algorithms

This paper introduces the Better Hide Communities (BHC) benchmark dataset aimed at standardizing evaluations in community deception across networks. BHC addresses the need for a common framework to assess the effectiveness of existing and perspective deception strategies by enabling their comparative analyses. BHC serves as a foundation for future work in developing sophisticated algorithms for community deception, enhancing the understanding of algorithmic abilities to employ deceptive measures within communities. Additionally, it offers valuable insights into the varying degrees of resilience that different detection algorithms exhibit against deception strategies.

Deceitful autocratization: Subverting democracy through electoral reform in Mexico

AbstractAntipluralist incumbents often resort to opaque and deceptive strategies to subvert democracy. One such strategy involves devising formal institutions that pretend to promote and deepen democracy while enforcing authoritarian rule instead. This study labels such a strategy as “deceitful autocratization” and argues that a case that illustrates how antipluralist presidents promote it is the electoral reform launched by Andrés Manuel López Obrador in Mexico. The analysis shows that López Obrador is a populist leader who also features prominent illiberal and antipluralist attributes. The article examines the political process that led to the legal modifications he introduced, which aimed at an overhaul of the country's system of electoral governance, one of the crucial foundations of Mexican democracy. The reform was supposed to democratize the regime, eliminate electoral fraud, and reduce the cost of elections. In fact, it represented a step forward in establishing competitive authoritarianism.

Global misinformation trends: Commonalities and differences in topics, sources of falsehoods, and deception strategies across eight countries

In a quantitative content analysis of 3,154 debunking articles from 23 fact-checking organizations, this study examines global misinformation trends and regional nuances across eight countries in Europe and Latin America (UK, DE, PT, SP, AR, BR, CL, and VZ). It strives to elucidate commonalities and differences based on political and media system indicators. Notably, countries with a substantial online presence of far-right parties avoid disclosing (fake) ordinary accounts to evade engaging in inauthentic coordinated actions. While entirely fabricated stories are infrequent, they stand out in Brazil and Spain, the two countries with higher political polarization. Despite variations, aggregated forms of fabrication (invented, manipulated, imposter, or decontextualized content) are more prominent in Latin America due to high social media use for news and low reliance on public media. Conversely, in Europe, countries are more impacted by misleading (cherry-picked, exaggerated, and twisted) information.

CGAN-based cyber deception framework against reconnaissance attacks in ICS

In recent years, Industrial Control Systems (ICSs) have faced increasing vulnerability to cyber attacks due to their integration with the Internet. Despite efforts to enhance cybersecurity, reconnaissance attacks remain a significant threat, prompting the need for innovative defensive strategies. This paper introduces a novel approach to strengthen the defensive capabilities of ICS networks against reconnaissance attacks using machine learning-driven cyber deception techniques. Leveraging Conditional Generative Adversarial Networks (CGANs), the proposed framework dynamically generates defensive network topologies to network shuffling and implement deception strategies, prioritizing system availability. Extensive simulations demonstrate the superior efficacy of the proposed framework in enhancing cybersecurity while minimizing computational overhead. By effectively mitigating reconnaissance attacks, this solution reinforces the resilience of ICS networks, safeguarding critical industrial infrastructure from evolving cyber threats. These findings underscore the significance of adopting machine learning-based cyber deception as a pragmatic security measure for protecting ICS networks in real-world industrial contexts.

NOTHING NEW IN MARKETING COMMUNICATION OF E-CIGARETTES – THE NEW WAVE OF FAKE NEWS

The tobacco industry is using the same methods of persuasion as they were doing since the start of advertising cigarettes and other tobacco products in 1914. Producers of tobacco products have been accused of spreading misinformation and fake news since the 1980. Main objective of our paper is to describe the evolution of “fake” marketing communication, especially of e-cigarettes - lately the deceptive strategies of e-tobacco companies in their advertising campaigns, where they minimize the dangers of smoking and portray it as a beneficial or attractive practice are coming under the control of the relevant institutions. Since the e-tobacco companies are using all the modern means and platforms of communication, especially young generations are quite successfully targeted. Results of desk research will be presented as well as some of the activities performed by state owned, and private institutions aimed to prevent harmful and shameless spreading of this conflicting information.

Deception Strategies in Occupation Discourse: An Analysis of Napoleon Bonaparte’s Discourse during the French Campaign in Egypt (1798-1801)

Deception Strategies in Occupation Discourse: An Analysis of Napoleon Bonaparte’s Discourse during the French Campaign in Egypt (1798-1801)

Deteksi Email Spam dengan Continuous Bag-Of-Words dan Random Forest

Spam email poses a significant cyber threat, as scammers employ various tactics to deceive individuals into divulging sensitive information or downloading harmful content. For instance, in June 2023, Indonesia encountered approximately 6.51 thousand spam attacks, underscoring the widespread nature of this issue. These attacks frequently involve deceptive strategies, such as impersonation or false promises of rewards, to ensnare unsuspecting victims. Succumbing to spam can result in financial losses and other grave repercussions. To address this concern, this research addresses this pressing problem by focusing on email content classification to detect phishing attempts. The proposed solution leverages runtime platforms such as Google Colab and uses Continuous Bag of Words (CBOW) analysis and Random Forest methods. CBOW is selected for its effectiveness in capturing semantic relationships between words, allowing the model to extract meaningful features from the email content. Random Forest, on the other hand, is chosen for its ability to handle imbalanced datasets commonly encountered in email classification tasks, ensuring fair representation of both spam and ham emails during model training. By combining these two techniques, we aim to develop a robust classification model capable of accurately distinguishing between phishing (spam) and legitimate (ham) emails, thus enhancing email security measures. Through our approach, we aim to classify the SpamAssassin dataset into ham or spam categories, with an anticipated precision rate of 0.98, demonstrating the model's effectiveness in accurately identifying phishing emails.

Deception abilities emerged in large language models

Large language models (LLMs) are currently at the forefront of intertwining AI systems with human communication and everyday life. Thus, aligning them with human values is of great importance. However, given the steady increase in reasoning abilities, future LLMs are under suspicion of becoming able to deceive human operators and utilizing this ability to bypass monitoring efforts. As a prerequisite to this, LLMs need to possess a conceptual understanding of deception strategies. This study reveals that such strategies emerged in state-of-the-art LLMs, but were nonexistent in earlier LLMs. We conduct a series of experiments showing that state-of-the-art LLMs are able to understand and induce false beliefs in other agents, that their performance in complex deception scenarios can be amplified utilizing chain-of-thought reasoning, and that eliciting Machiavellianism in LLMs can trigger misaligned deceptive behavior. GPT-4, for instance, exhibits deceptive behavior in simple test scenarios 99.16% of the time (P < 0.001). In complex second-order deception test scenarios where the aim is to mislead someone who expects to be deceived, GPT-4 resorts to deceptive behavior 71.46% of the time (P < 0.001) when augmented with chain-of-thought reasoning. In sum, revealing hitherto unknown machine behavior in LLMs, our study contributes to the nascent field of machine psychology.

Chemical imitation of yeast fermentation by the drosophilid-pollinated deceptive trap-flower Aristolochia baetica (Aristolochiaceae)

Deceptive flowers, unlike in mutualistic pollination systems, mislead their pollinators by advertising rewards which ultimately are not provided. Although our understanding of deceptive pollination systems increased in recent years, the attractive signals and deceptive strategies in the majority of species remain unknown. This is also true for the genus Aristolochia, famous for its deceptive and fly-pollinated trap flowers. Representatives of this genus were generally assumed to be oviposition-site mimics, imitating vertebrate carrion or mushrooms. However, recent studies found a broader spectrum of strategies, including kleptomyiophily and imitation of invertebrate carrion. A different deceptive strategy is presented here for the western Mediterranean Aristolochia baetica L. We found that this species is mostly pollinated by drosophilid flies (Drosophilidae, mostly Drosophila spp.), which typically feed on fermenting fruit infested by yeasts. The flowers of A. baetica emitted mostly typical yeast volatiles, predominantly the aliphatic compounds acetoin and 2,3-butandiol, and derived acetates, as well as the aromatic compound 2-phenylethanol. Analyses of the absolute configurations of the chiral volatiles revealed weakly (acetoin, 2,3-butanediol) to strongly (mono- and diacetates) biased stereoisomer-ratios. Electrophysiological (GC-EAD) experiments and lab bioassays demonstrated that most of the floral volatiles, although not all stereoisomers of chiral compounds, were physiologically active and attractive in drosophilid pollinators; a synthetic mixture thereof successfully attracted them in field and lab bioassays. We conclude that A. baetica chemically mimics yeast fermentation to deceive its pollinators. This deceptive strategy (scent chemistry, pollinators, trapping function) is also known from more distantly related plants, such as Arum palaestinum Boiss. (Araceae) and Ceropegia spp. (Apocynaceae), suggesting convergent evolution. In contrast to other studies working on floral scents in plants imitating breeding sites, the present study considered the absolute configuration of chiral compounds.

Ecological perspectives on female and male reproductive success with competition in two Serapias species.

The deceptive strategies by which orchids are pollinated and how these are capable of attracting pollinators remain understudied with regard to their implications for plant fitness. Despite their ecological importance, limited investigations have been conducted on sexual deception and shelter mimicry in orchid species, making this a compelling avenue in orchid biology research. To expand the knowledge of these reproductive mechanisms, we studied the pollination of Serapias lingua and S. parviflora in co-occurring and isolated sites in the Balearic Islands (Spain), further accentuated by the presence of a hybrid, indicating shared pollinators. We employed bagging and hand pollination experiments to examine the reproductive biology of the two species. Furthermore, we evaluated the influence of phenotypical and ecological factors on reproductive success, including biometric measurements, reproductive performance and neighbourhood diversity (co-flowering and pollinator communities). Reproductive mechanisms between these two orchid species exhibit substantial disparities. Serapias lingua relies primarily on insect-mediated pollination, while S. parviflora demonstrates self-reproduction capacity. Although events of open pollination are rare, hybridization occurs predominantly when S. lingua is the pollen donor. Fruit set in S. parviflora was positively correlated with plant height, while in S. lingua it was negatively associated with flower size. The coexistence of the two species positively affected pollinium removal in S. parviflora, but did not exert an influence on reproductive traits in S. lingua. Overall, biometric parameters were diminished in isolated compared with co-occurring sites. At the community level, the increased diversity of co-flowering species in the vicinity exhibited an inhibitory effect on pollinium removal in S. parviflora. Under a context of pollinator loss or phenological mismatch between pollinator presence and flowering, the selfing capacity of S. parviflora would guarantee reproduction whereas S. lingua survival would be compromised. Furthers studies are needed to assess the effects of phenotypical and ecological factors on reproductive success of S. lingua in pollinator-decline scenarios.

Awareness and Attitude Towards Tobacco Products and Tobacco Industry and Perception about Government's Role in Tobacco Control among the Adult Rural Population: A Cross-sectional Study in the Indian State of Bihar.

Attitudes and impressions toward the tobacco industry and tobacco products among the general public are important determinants for curbing the menace of the tobacco epidemic. Accordingly, this study aimed to assess the knowledge and perceptions about the tobacco industry and tobacco products and analyze attitudes towards social denormalization (SD) of tobacco use and tobacco industry denormalization (TID) among the rural population of Bihar, India. This community-based cross-sectional survey was conducted on 421 adults aged 18 to 65 years who were selected using multistage systematic random sampling in a rural area of Bihar State in India from January to March 2022. Results were presented as proportions and the factors associated with support for TID and SD were identified using the chi-square test and binary logistic regression. Out of 421 participants, 342 (81.2%) did not consider smokeless tobacco to be very dangerous. Nearly half (192, 45.6%) of the individuals believed that tobacco companies never tell the truth about the ill effects of tobacco use on health. Maximum, 345 (89.5%) also believed that the tobacco industry is responsible for adverse health effects of tobacco use and that the government should sue them. The prevalence of favorable attitudes toward TID and SD was found to be 55.1% [95% CI: 50.3% - 59.8%] and 38.2% [95% CI: 33.7% - 42.9%], respectively. One out of every two and one out of every three individuals showed favorable attitudes toward TID and SD, respectively. There is a need to inform and educate the public on the ill effects of tobacco and the deceptive strategies used by the tobacco industry to help them choose health over tobacco.

What are the drivers of female success in food-deceptive orchids?

A large suite of floral signals, and environmental and biotic characteristics influence the behavior of pollinators, affecting the female success of food-deceptive orchids. In this study, we examined the many factors shaping the reproductive output of three orchid taxa: Dactylorhiza majalis, D. incarnata var. incarnata, and D. fuchsii. We applied a statistical model to correlate female success (number of fruit sets) with individual characteristics (plant and inflorescence height, number of flowers, and spur length), number of pollinaria removed, flowering time, and density of floral units of co-flowering rewarding plants. Our findings suggested that the broad spectrum of variations in Dactylorhiza's morphological traits, floral display, and flowering phenology within different environmental contexts has a significant impact on their reproductive success. The number of fruits increased with an increase in the number of pollinaria removed in the studied Dactylorhiza taxa. In contrast, a higher number of flowers per inflorescence and higher inflorescences in relation to individual height always decreased fruit set. We observed that low number of co-flowering rewarding plants in populations could affect the Dactylorhiza reproductive output as magnets and competitor plants. The synchronization of flowering, or lack thereof, between Dactylorhiza and rewarding plants can limit reproductive success. This demonstrates that the food deception strategy is multidirectional, and reproductive output can vary considerably both spatially and temporally within the context of this strategy.

Deceptive learning in histopathology.

Deep learning holds immense potential for histopathology, automating tasks that are simple for expert pathologists and revealing novel biology for tasks that were previously considered difficult or impossible to solve by eye alone. However, the extent to which the visual strategies learned by deep learning models in histopathological analysis are trustworthy or not has yet to be systematically analysed. Here, we systematically evaluate deep neural networks (DNNs) trained for histopathological analysis in order to understand if their learned strategies are trustworthy or deceptive. We trained a variety of DNNs on a novel data set of 221 whole-slide images (WSIs) from lung adenocarcinoma patients, and evaluated their effectiveness at (1) molecular profiling of KRAS versus EGFR mutations, (2) determining the primary tissue of a tumour and (3) tumour detection. While DNNs achieved above-chance performance on molecular profiling, they did so by exploiting correlations between histological subtypes and mutations, and failed to generalise to a challenging test set obtained through laser capture microdissection (LCM). In contrast, DNNs learned robust and trustworthy strategies for determining the primary tissue of a tumour as well as detecting and localising tumours in tissue. Our work demonstrates that DNNs hold immense promise for aiding pathologists in analysing tissue. However, they are also capable of achieving seemingly strong performance by learning deceptive strategies that leverage spurious correlations, and are ultimately unsuitable for research or clinical work. The framework we propose for model evaluation and interpretation is an important step towards developing reliable automated systems for histopathological analysis.