Decision-Making in the Age of AI: A Review of Theoretical Frameworks, Computational Tools, and Human-Machine Collaboration

“By training on extensive datasets, these networks enable highly accurate visual-based decision-making, leading to revolutions in self-directed vehicles, reconnaissance structures, and medicinal copy investigation [70].”

“According to prospect theory, individuals interpret outcomes in terms of gains and losses, with these perceptions shaped by how the options are framed and the decision-maker’s expectations [21].”

“More advanced RNN variants, such as long short-term memory (LSTM) [57] along with gated recurrent units (GRU) [58] further enhance this ability by effectively managing long-term dependencies and overcoming the vanishing gradient problem [59].”

“…learning (RL) networks, which conglomerate neural networks with RL algorithms, enable systems to learn over environmental interactions, making them particularly effective for decision-making chores involving tardy recompenses, such as those in gaming, robotics, and autonomous systems [61].”

“More advanced RNN variants, such as long short-term memory (LSTM) [57] along with gated recurrent units (GRU) [58] further enhance this ability by effectively managing long-term dependencies and overcoming the vanishing gradient problem [59].”

“E-commerce platforms, streaming services, and social media networks leverage these models to enhance user experience, boost engagement, and optimize content delivery efficiency [76, 77].”

“Neural networks have demonstrated strong performance in tasks such as language translation, sentiment analysis, and question-answering [72–74].”

“More advanced RNN variants, such as long short-term memory (LSTM) [57] along with gated recurrent units (GRU) [58] further enhance this ability by effectively managing long-term dependencies and overcoming the vanishing gradient problem [59].”

“This idea is captured in the Dual Process Concept of cognizance, spread by Daniel Kahneman [28], illustrated in Figure 7, which distinguishes between two modes of thinking: System 1 is characterized by speed, intuition, and automatic processing, while System 2 operates in a slower, more deliberate,…”

“RNNs, which are planned to handle sequential data such as time series and natural language, are well-suited for modelling temporal dependencies thanks to their recurrent connections [56].”

“Heuristics operate within a fast, intuitive, often automatic, and subconscious information processing system, contrasting with a slower, more deliberate, accurate, and effort-intensive system [27].”

“As a result, deep learning has significantly enhanced human-computer interactions, transforming how individuals engage with technology [71, 72].”

“Foundational architectures in deep learning include CNNs, multilayer perceptron, restricted Boltzmann machines (RBMs), auto-encoders, and RNNs [62–64].”

“Researchers emphasize that humans are more adept in areas such as judgment, political analysis, psychological insight, Agility, imagination, strategic foresight, along with managing ambiguity [119, 121, 122, 125, 126, 129].”

“…AI’s role in gathering information and monitoring status within production and logistics, utilizing tools like the data science toolbox by Flath et al. [99], supply chain risk management tools by Baryannis et al. [94, 97], and the self-adaptive supply chain projected by Calatayud et al. [100].”

“However, there is no consensus on whether AI will ever fully replicate human implicit abilities [121, 136].”

“Furthermore, AI cannot fully replace the advantages of human group decision-making and may, in certain instances, intensify the complexities involved in being decision-making processes [99, 117].”

“By identifying patterns within large datasets, these models facilitate dynamic decision-making, helping to mitigate risks and improve financial performance [79].”

“Deep learning, particularly through neural networks, has revolutionized numerous fields by enabling the processing of large datasets and the identification of complex patterns [68].”

“To address this, frameworks such as AI Impartiality 360 [118] as well as Open Algorithm [110] offer preprocessing, inprocessing, as well as post-processing techniques aimed at reducing bias, though not eliminating it entirely.”

“Recent advancements in decision-making and artificial intelligence have been discussed extensively in various studies [9, 10].”

“Moreover, research demonstrates that continuous training and hands-on experience with AI foster trust in the technology, thereby enhancing its overall efficacy [110, 111].”

“Researchers emphasize that AI requires well-defined objectives, as it struggles with managing uncertainty and processing complex inputs [102, 119, 122].”

“Recurrent neural networks (RNNs), featuring recurrent connections, excel in handling sequential data by modeling temporal dependencies [53].”

“Their use has been instrumental in reducing costs, assigning resources, and optimizing routes [34].”

“Neural networks consist of interconnected neurons arranged in layers-typically comprising input, hidden, and output layers [46].”

“These theories often account for cognitive biases and heuristics, which are incorporated into AIbased models, particularly neural networks, and computational tools, to optimize decision-making processes [15].”

“Bienhaus et al. [103] and Butner et al. [104], drawing from survey data, advocate for a complete process redesign instead of merely adding new technologies onto existing frameworks.”

“Scholars including Migliore et al. [112], Bader et al. [109], and Whittle et al. [113] advise assessing the skills employees need to work effectively with AI, with leadership providing guidance during this transition based on their AI expertise [111, 113, 115].”

“MCDM methods in such cases are often called compensatory methods [29], where trade-offs between criteria are allowed.”

“Nilsson [133] notes that AI can only “interact with foresight in its environment” when directed by human input, a view contrasting with Simon’s [96] comparison of AI and human capabilities.”

“These algorithms have become invaluable for analyzing and gaining intuitions from this intricate digital environment, commonly known as the “echoverse”, comprising user content, organizational material, traditional news, and press releases [67].”

“Paschen et al. [105] support this transition by offering a four-dimensional framework to assess if AI adoption fosters novelty in produces or else progressions and whether it enhances or reduces human competencies.”

“E-commerce platforms, streaming services, and social media networks leverage these models to enhance user experience, boost engagement, and optimize content delivery efficiency [76, 77].”

“The literature outlines several practical and hypothetical applications of AI, particularly in data interpretation, generating alternatives, and establishing probabilities and preferences, which may also include assessing potential consequences [94, 97, 98].”

“Normative theories present ideal decision-making processes based on rational calculations, while descriptive theories focus on explaining how people make decisions in the real world [13, 14].”

“While human judgment is still considered essential for utility assessments [98], AI can assist by forecasting the potential impact of various decision alternatives on the organization and its stakeholders [94, 101, 126].”

“To enhance the validity and reliability of our review findings, we utilized a systematic literature review approach informed by Pittway’s [1] ideologies: pellucidity, precision, assimilation, emphasis, impartiality, approachability, and handling.”

“More recent studies have emphasized the groundbreaking exploration of contrasting paradigms between normative and descriptive decision-making theories [11, 12].”

“By integrating MCDM with AI-driven optimization, decision-makers can achieve more accurate, flexible, and data-driven outcomes, bridging the gap between traditional decision analysis and modern computational intelligence [30, 31].”

“Such tendencies may be influenced by cognitive biases (Pronin), which have been observed in various species [22].”

“Nilsson [133] notes that AI can only “interact with foresight in its environment” when directed by human input, a view contrasting with Simon’s [96] comparison of AI and human capabilities.”

“To enhance AI literacy, researchers emphasize involving both top management and impacted employees in AI initiatives, given that acceptance varies across organizational levels [111, 114].”

“…AI’s role in gathering information and monitoring status within production and logistics, utilizing tools like the data science toolbox by Flath et al. [99], supply chain risk management tools by Baryannis et al. [94, 97], and the self-adaptive supply chain projected by Calatayud et al. [100].”

“As a result, “while humans have access to both explicit and tacit knowledge, AI is limited by its lack of access to tacit knowledge and its reliance on historical data from which patterns are identified” [91].”

“Researchers agree that AI can be effectively employed to collect, interpret, evaluate, in addition to propagate information, enhancing the speed, volume, diversity, and availability of data [88–90].”

“Convolutional neural networks (CNNs), frequently employed in image in addition to video processing, are designed to capture spatial hierarchies through convolutional layers [51, 52].”

“In the financial sector, deep learning enhances socially answerable investing and assortment optimality, helping align investment strategies with ethical and sustainability principles [86].”

“Neural networks excel at analyzing vast amounts of complex financial data, forecasting stock market trends, detecting fraudulent transactions, and supporting algorithmic trading and credit risk evaluation [78].”

“The selection of AI applications is influenced by an organization’s structure and resource allocation, although studies indicate that AI can, in turn, impact organizational design and structure [107, 108].”

“Foundational architectures in deep learning include CNNs, multilayer perceptron, restricted Boltzmann machines (RBMs), auto-encoders, and RNNs [62–64].”

“One of the most significant impacts has been in computer vision, where deep learning has redefined fundamental tasks such as object detection, facial recognition, and image classification, ushering in a new era of advancements [69].”

“At present, AI contributes most significantly to the input and processing stages, particularly within knowledge management [89, 95, 102, 132].”

“Decision-making, in this context, entails choosing from a range of possible actions [16].”

“Meanwhile, in highway infrastructure management, deep learning-based automated decision-making systems improve maintenance prioritization and scheduling, streamlining operations and enhancing long-term decision effectiveness [83].”

“The aim is to learn how to make optimal choices by developing algebraic models of preferences grounded in idealized behavioral principles [6].”

“The selection of AI applications is influenced by an organization’s structure and resource allocation, although studies indicate that AI can, in turn, impact organizational design and structure [107, 108].”

“Applications of deep learning have also seen extensive use in the burgeoning field of social media, resulting in massive quantities of unstructured, user-produced media, including text, images, and videos [66].”

“More recent studies have emphasized the groundbreaking exploration of contrasting paradigms between normative and descriptive decision-making theories [11, 12].”

“Numerous studies highlight the complementary relationship between humans and AI, where AI systems improve through human input and humans gain insights from AI support [119, 123].”

“The issue of moral responsibility in AI decision-making remains unresolved, as it is currently impossible to embed comprehensive moral principles into algorithms [134, 135].”

“Scholars including Migliore et al. [112], Bader et al. [109], and Whittle et al. [113] advise assessing the skills employees need to work effectively with AI, with leadership providing guidance during this transition based on their AI expertise [111, 113, 115].”

“CNNs excel in image linked chores-including arrangement, item discovery, and dissection due to their capability to efficiently process visual data [55].”

“Over the past few years, deep learning has been rapidly adopted by managers to improve decision-making processes across various organizational levels [65].”

“While human judgment is still considered essential for utility assessments [98], AI can assist by forecasting the potential impact of various decision alternatives on the organization and its stakeholders [94, 101, 126].”

“…AI autonomy viable for certain decisions, they also stress that for high-stakes decisions, human oversight or “veto power” is essential [125, 127], highlight AI’s ability to “automate tasks”, allowing humans to focus on higher-value work, while Klumpp et al. [128] advise that AI shifts…”

“…decisions, human oversight or “veto power” is essential [125, 127], highlight AI’s ability to “automate tasks”, allowing humans to focus on higher-value work, while Klumpp et al. [128] advise that AI shifts human roles toward supervisory responsibilities rather than direct task execution.”

“Terziyan et al. [92] propose the most comprehensive approach for incorporating diverse types of information into decision-making, their patented Pi-Mind methodology seeks to replicate human decision-makers by capturing soft facts and potential utility levels.”

“Normative theories present ideal decision-making processes based on rational calculations, while descriptive theories focus on explaining how people make decisions in the real world [13, 14].”

“Convolutional neural networks (CNNs), frequently employed in image in addition to video processing, are designed to capture spatial hierarchies through convolutional layers [51, 52].”

“For instance, Costache et al. [85] combined deep learning and fuzzy logic, in addition to MCDM investigation to address flash flood risks.”

“Various neural network architectures are utilized in different areas of decision-making devel-opment [47–50].”

“Neural networks have demonstrated strong performance in tasks such as language translation, sentiment analysis, and question-answering [72–74].”

“Furthermore, AI cannot fully replace the advantages of human group decision-making and may, in certain instances, intensify the complexities involved in being decision-making processes [99, 117].”

“Consequently, humans rely on heuristics-mental shortcuts or “rules of thumb”-to facilitate decision-making [23, 24].”

“Various neural network architectures are utilized in different areas of decision-making devel-opment [47–50].”

“Researchers emphasize that humans are more adept in areas such as judgment, political analysis, psychological insight, Agility, imagination, strategic foresight, along with managing ambiguity [119, 121, 122, 125, 126, 129].”

“Researchers commonly acknowledge that AI can gather large volumes of evidence regarding varied resources, streamline data allocation, in addition to it assist in construal, enhancing knowledge management in terms of efficiency and speed [88, 90, 95, 104].”

“Studies highlight those machines cannot be held morally accountable, underscoring the necessity of adapting ethical constructs, such as guilt and fairness, to AI contexts [136].”

“At present, AI contributes most significantly to the input and processing stages, particularly within knowledge management [89, 95, 102, 132].”

“The FNN, also known as a multilayer perceptron, is the most basic neural network architecture and is widely applied to tasks such as classification, regression, and pattern recognition [54].”

“Various frameworks have been proposed to allocate tasks between humans and AI, spanning from fully automated AI solutions to hybrid models and human-exclusive decisions [121, 124].”

“Similarly, in the energy sector, intelligent energy management systems utilize long short-term memory (LSTM) models alongside fuzzy decision-making approaches to optimize energy distribution, improving overall efficiency and sustainability [84].”

“Some scholars advocate for advancing AI into more sophisticated, self-learning systems with greater intelligence [115, 125, 126, 137].”

“They form the basis for reshaping or establishing new organizational frameworks necessary for smooth AI integration [102, 105, 106].”

“Normative guidelines serve as rational benchmarks to assess actual behavior [7], helping individuals maximize the expected utility of their choices [8].”

“As Rousseau [131] points out, “recognizing biases in others is often easier than seeing them in ourselves”.”

“The issue of moral responsibility in AI decision-making remains unresolved, as it is currently impossible to embed comprehensive moral principles into algorithms [134, 135].”

“However, Parry et al. [125] and Agrawal et al. [126] consider complete AI autonomy viable for certain decisions, they also stress that for high-stakes decisions, human oversight or “veto power” is essential [125, 127], highlight AI’s ability to “automate tasks”, allowing humans to focus on…”

“Incorporation of CNNs with fuzzy logic further enhances analytical reliability, enabling more cognizant decision-making in complex healthcare scenarios [87].”

“Effective and responsible supervision of AI systems demands a comprehensive grasp of both the technology’s limitations and its potential [109, 113, 130].”

“In medical diagnostics, convolutional neural networks (CNNs) have proven highly effective in delivering precise assessments, aiding critical decision making in medical treatment [80].”

“By integrating MCDM with AI-driven optimization, decision-makers can achieve more accurate, flexible, and data-driven outcomes, bridging the gap between traditional decision analysis and modern computational intelligence [30, 31].”

“Bienhaus et al. [103] and Butner et al. [104], drawing from survey data, advocate for a complete process redesign instead of merely adding new technologies onto existing frameworks.”

“MADM techniques are particularly vital in fields such as management and engineering, where decision-makers must choose the most suitable option from a limited pool of alternatives [32].”

“AI literacy broadly encompasses a comprehensive understanding of AI’s strengths and limitations-a competency often lacking in organizations, as Whittle et al. [113] observe.”

“Researchers emphasize that AI requires well-defined objectives, as it struggles with managing uncertainty and processing complex inputs [102, 119, 122].”

“While human judgment is still considered essential for utility assessments [98], AI can assist by forecasting the potential impact of various decision alternatives on the organization and its stakeholders [94, 101, 126].”

“In other words, decision-making in certain situations requires rarer cognitive effort than the ideologies outlined in Expected Utility Theory suggest [17].”

“These models power essential technologies, including chatbots, virtual assistants, and search engines, allowing for more sophisticated interpretation and response to textual data [75].”

“Cheng et al. [82] applied CNNs to enable data-driven decision-making in autonomous systems, promoting safer and more effective transport solutions.”

“Furthermore, integrating pattern recognition tools like PANN significantly augments the precision and effectiveness of medicinal decision-making [81].”

“In recent years, as social application scenarios have grown more complex, traditional static frameworks have revealed their limitations, and conventional CD theories have shown noticeable shortcomings [11].”

“The researchers used this deep auto-encoder to reduce the dimension of the mammogram images without reducing any important discriminating information [16].”