A lot of research has shown the extent to which trust and AI systems following moral rules are linked.⁹ This is very important for keeping users’ information safe and allowing everyone to use the system. Stahl¹⁰ found that people are more likely to trust AI if they think it is moral. This is especially true when it comes to privacy and its functioning. It turns out that moral behaviour has a big impact on how much people trust a website.¹¹ When people and businesses use AI systems, one of the main things they are concerned about is the safety of user data. When privacy is shattered, bigger things can happen in society and the law, resulting in a lower level of trust.

Open AI systems make people more likely to trust them. With open algorithms, there is less of the unknown that comes with black-box algorithms.¹² Now, though, a lot of algorithms are very hard to grasp, which makes it difficult to make AI more open. A lot of the time, they have to work with very large datasets and machine learning models that are tough for non-experts to fully grasp. Thakur et al¹³ assert that trust is important for AI to work well in fields with a lot of data, such as healthcare, e-commerce, and finance. They discovered that people are more likely to trust AI systems when there are strong safeguards in place to ensure that data is handled fairly. In these frameworks, rules tell us how to get, store, use, and share the data. Mazurek and Małagocka¹⁴ claim that rules that make AI systems follow data privacy laws (like General Data Protection Regulation [GDPR]) and make it clear how user data is used can help people trust these systems.

To ensure that AI is used in a way that is acceptable to society, it is morally important to adhere to privacy regulations, be transparent, be useful, and reduce bias (as shown in Figure 1). The most important things to think about when it comes to the law are accountability, privacy, responsibility, correctness, and freedom of choice. Safety, cognitive services, accountability, acceptance, and robot rights are just a few of the many interconnected big ideas. It is important to find a balance between legal and moral issues so that AI systems are trustworthy, open to everyone, and responsible.

It can be hard to trust AI because it is always changing and getting smarter. Therefore, static or binary models cannot be used to check whether it is reliable. AI that follows rules is often made more ethical by adding rule-based systems. According to studies, rule-based systems are overly inflexible to handle the ambiguities and shifts that occur when making obvious moral decisions.¹⁶ One example of a rule-based system is the idea that AI should always put the privacy of data first. But in real life, getting privacy might mean that things are less open or fair.^17,18 When AI systems are too fixed in their ways, it can be hard for them to make fair decisions that take everything into account.

“Ethics by Design” in AI means taking moral, legal, and technological issues into account to make sure that AI development is done in a responsible way.²⁰ A big part of “Ethics by Design” is making sure that AI systems have morals from the very beginning. Around this core, we can see three main areas: law, ethics, and technology, as depicted in Figure 2.²¹ Law experts emphasize that AI systems should be limited in what they can do, their intellectual property (IP) should be protected, and they should be given tools to assist them in following the rules. The last domain is Technology, which looks at what can be done technically. It examines tools that can help make rules and protect “Onlife” rights, which are rights that cover both online and offline interactions.²² While AI is progressing, these moral and legal limits need to be taken into account. Morality is subjective and up to individual judgement. It speculates on how AI, once implanted in human brains, might one day become self-aware, as AI has the potential for both good and evil, and AI fundamentally lacks moral or ethical dimensions.²³ Elon Musk considers it the most important technological advance of our time.

In fuzzy logic, there are different levels of truth, not just true or false options like in binary logic.²⁴ For this reason, it is very good at modelling uncertainty and slow changes in situations where choices are not clear.^25,26 The groundbreaking work that Zadeh did on fuzzy logic gives us a way to deal with not knowing what to do or being too vague. Because of this, it is perfect for modelling trust, which is affected by many variables that are not always easy to measure. The level of trust that people have in AI systems can be influenced by how effectively the systems adhere to moral principles, such as privacy and fairness. When people’s trust levels rise or fall gradually rather than suddenly, fuzzy logic can model this phenomenon.

Some new research has looked into how fuzzy logic can be used in AI, especially when the AI has to make moral decisions. For their study, Alonso et al²⁷ used fuzzy logic to make a model of the moral problems faced by AI systems. What they showed is that AI can make better decisions with the help of fuzzy rule-based systems. Some ethical dimensions are privacy, fairness, and transparency. By giving these dimensions fuzzy values, the AI system can compare them and make decisions that are not either/or but instead show a balanced ethical approach. It is important to be able to deal with trade-offs and ambiguities when there is not a single answer that is completely moral in every way.

The digital economy makes it even more important to have good data governance if we want people to trust AI systems.^28,29 The economy we have now makes data more valuable, so the systems that gather, store, and use it need to make sure that AI systems follow privacy laws and are honest. A study by Saura et al.,³⁰ proclaims that to build and keep trust, there is a need for strong data governance. This is very important in fields where private data is handled by AI, like healthcare and finance.

Most of the time, data governance frameworks have rules about how to handle user data. These rules include privacy policies, moral guidelines, and ways to make sure that rules are followed.³¹ Setting up these kinds of systems can change the way trust works directly. When people know AI will be honest and safe with their data, they are more likely to trust it. Frameworks for data governance should be adaptable since rigid models might not be able to keep up with the changes that happen over time in AI applications that are driven by data.^32,33

Finding the best balance between moral issues like privacy, fairness, and openness is a big part of data management. For this reason, AI systems need to be able to make moral decisions in real time so that the system keeps running smoothly. Because it allows AI to think about moral issues as fuzzy variables, fuzzy logic is believed by many to be a potential solution to this problem. However, AI can assess the degree to which fairness, privacy, and transparency are upheld without actually deciding on these areas.

In a fuzzy logic-based system, parts of ethics like fairness, privacy, and openness are given fuzzy membership functions.^34,35 We can see how well these dimensions are met by these functions. One example of a value for privacy is 0 (no privacy) and 1 (complete privacy). Values in the middle show how well privacy rules are being followed in some situations. After that, systems with fuzzy rules can be built to help the AI make moral decisions. The ethical decision score will be high if all of these things are high: privacy, fairness, and transparency. If one or more of these things is low, the ethical decision score could also be moderate or low. By utilizing fuzzy logic, AI can make more complex moral judgements that weigh the pros and cons of various moral dilemmas. It is possible to fully understand how trust changes in AI systems when we look at both fuzzy trust modelling and fuzzy ethical decision-making together. In the digital economy, AI will play a bigger role, so it will be important to be able to model trust as a variable that changes over time.

The main variables chosen in this fuzzy logic model are privacy, fairness, transparency, and trust because they play such an important role in ethical AI governance. These variables show important aspects of following the rules of ethics that have a big effect on how much people trust AI systems. The values of each variable are represented as fuzzy variables, which means they can be any number between 0 and 1. They are not limited to two states, like “privacy-compliant” or “not privacy-compliant.” This setup lets the model show how ethical behaviour changes over time, so it can adapt to the complex realities of AI ethics. Variables’ description is presented in Table 1.

These variables were chosen to show important ethical issues that influence people’s decision to trust AI systems. The model can change trust scores based on the exact mix of privacy, fairness, and transparency levels because it models each ethical dimension separately.

By using these membership functions, the fuzzy logic model can capture the subtleties of ethical adherence in AI systems, providing a trust score that realistically reflects how varying degrees of privacy, fairness, and transparency influence overall trust. This flexible approach allows the model to adapt to different contexts, making it a robust tool for evaluating trust in ethical AI governance.

Data governance frameworks that use AI systems illustrate the operation of trust and the making of moral decisions using fuzzy logic. Our method is made up of two main parts: a fuzzy rule-based system for deciding what is right and wrong and fuzzy trust levels. As a whole, it should demonstrate how challenging it is to trust another person while still acting morally. Because of these updates, the AI system trust model is much better now as it becomes both more adaptable and accurate. AI systems use a fuzzy variable () to represent trust. Its value can be any number between 0 (totally not trust) and 1 (totally trust). It changes over time based on how the AI system works. When it comes to trust, being honest, open, and respectful of privacy are all important moral actions.

Membership Functions and Their Application

In the fuzzy logic model, each variable has a membership function that tells the model how to map input values to a fuzzy level of adherence. The shapes of the membership function depend on how the variable affects trust and the level of ethical adherence. The sigmoid and triangular functions used in this model were picked because they can show gradual changes and clear thresholds, respectively (Table 2).

The sigmoid function is used for privacy and transparency because trust in these dimensions typically grows progressively with incremental improvements. For example, as privacy measures strengthen, user trust rises steadily, eventually levelling off once maximum adherence is achieved. This smooth increase aligns with real-world expectations where continuous compliance enhances trust.

To express trust levels mathematically, we employ a sigmoid membership function, μ_Trust(x), a widely used function in fuzzy logic to model gradual changes:

where X shows the AI system’s ethical behaviour score, which is a total score based on how well it follows ethical rules. k is a constant that controls how sharply trust increases with improved ethical behaviour. Higher values of k result in a steeper curve, which means that trust grows faster when certain moral standards are met. For instance, if an AI system only partially follows ethical rules, x would be worth very little, leading to a lower trust level. But as compliance goes up, like meeting important privacy standards, x gets closer to θ, which makes T(x) go up sharply as trust grows.

The triangular membership function is used for fairness as it has a clear peak value, indicating an ideal adherence range. In this case, the function assigns the highest trust level to a balanced, unbiased state, with declines on either side if fairness decreases. The triangular shape is ideal here because fairness often has a more discrete impact: Even slight biases can cause trust to drop sharply, which the triangular function captures.

For fairness (F), which typically has an optimal range, we use a triangular membership function:

where x is the fairness adherence level, a is the ideal fairness level at which the maximum trust is achieved, and b represents the tolerance range around the ideal fairness level. Values outside this range lead to a sharp decline in trust. This triangular function reflects that trust is highest when the fairness level is close to the ideal range (around a) and decreases sharply when fairness deviates from this optimal level. This models the reality that even slight biases can significantly reduce trust.

This study tests how well AI systems can make moral choices by using both a fuzzy variable to model trust and a fuzzy rule-based system. To make ethical choices in AI, we often have to balance a lot of different ethical principles, like fairness, privacy, and being open. Every one of these rules is shown as a fuzzy number that ranges from 0 to 1. If the value is greater than 0, people are following the rule. If the value is greater than 1, then everyone is following the rule to the letter.

These general things help us figure out how moral the AI is: It is important to keep user data safe. Safety (P) checks how well the AI system follows safety rules for privacy. The degree to which the system strives for impartiality in selecting winners and preventing bias from impacting results is indicated by the “fairness” (F) variable. How effectively an open system defends and explains its decisions to users and other important parties is one indicator of its openness. This is done with the transparency (Tr) variable. The overall score for moral behaviour goes up because of these moral factors. E(x) shows how well the AI system does in terms of being fair, private, and open.

The final ethical decision score E(x) is calculated using a weighted aggregation of the fuzzy variables P, F, and Tr:

where w₁, w₂, and w₃ are weights, and the trust model shows how important each ethical dimension is by giving it a score. The scores range from “somewhat” to “very important.” Firms can put different ethical issues at the top of their list based on the needs of their industry since these weights are adjustable. Concerns about the law and right and wrong may make privacy more important in healthcare. Some places, like schools, value openness more than others, so being able to explain decisions might be more important there.

There is one trustworthiness score made up of the fuzzy trust score and the ethical decision score. This is done to show how trust works in the AI system. This score shows how trustworthy the AI system is as a whole based on how well it follows moral rules and gains users’ trust over time. The overall trust score T_total is calculated by aggregating the ethical dimensions (privacy, fairness, and transparency) within the fuzzy framework:

This score lets us know how well AI systems keep our privacy safe, are fair, and are open. We can use fuzzy logic to show how AI systems trust each other and decide what is right or wrong on a small or large scale. This time, trust is shown as a range instead of a single number. This proves that trust grows over time and gives a way to handle rule-based moral binds. Privacy, fairness, and openness are valued in different ways in different fields, so this method can be used in those fields. Firms can change the way they check trust based on their own needs and the rules that affect them.

We use foundational theories and new empirical applications of fuzzy logic in ethical AI and decision-making frameworks to show that our fuzzy logic approach works. This process of validation makes sure that our model is strong and useful for studying how trust works in AI systems.

Zadeh⁶ came up with the basic ideas of fuzzy logic in 1965. Since then, there have been many improvements to the model, including his important work in 1996 that showed how useful it is in situations with uncertainty and slow changes. In his theory, which was based on the degrees of truth instead of absolute states, variables that do not fit into simple binary definitions could be modelled. Because of this, fuzzy logic is perfect for use in ethical AI, where strict moral standards, like protecting privacy or making sure everyone is treated fairly, cannot be properly shown by rigid, binary classifications. Fuzzy logic allows us to make more realistic and flexible decisions by consenting to choose from a range of values. This is especially helpful when we need to make ethical trade-offs based on the situation, which helps AI systems understand how trust works in complex ways.

Fuzzy logic is a great way to model trust in ethical AI because it takes into account how trust grows over time, which is something that binary or deterministic decision-making models cannot do well. Fuzzy logic is different from traditional binary logic because it does not just accept true or false statements. When AI systems are used in the real world, moral principles like privacy, fairness, and openness are not always absolute. Instead, they change levels of strictness depending on the situation, the level of compliance, and the expectations of stakeholders. Fuzzy logic is a more realistic and flexible way to judge trust because it lets people follow ethical rules in some situations.

When it comes to ethical AI, trust is always changing. Instead of going from “trusted” to “untrusted” all at once, it changes over time based on what an AI system does and how well it follows ethical rules. As an example, an AI app might slowly gain trust if it always follows privacy rules. On the other hand, a small privacy breach could lower trust without completely erasing it. Fuzzy logic can describe this changing pattern, where trust can rise or fall smoothly depending on how ethically sound someone is acting. This adaptability is very important for ethical AI because trust must be earned and kept by always following ethical rules. A model that only uses binary logic would miss these subtleties, giving a too-simplified picture of trust that could be wrong in more complicated situations.

Also, fuzzy logic works really well in situations where different morals might clash or need to be traded off. Ethical principles like fairness, privacy, and transparency do not always work well together in AI systems. For example, to improve transparency, some information might have to be made public, which could put privacy at risk. Fuzzy logic is a good way to deal with these kinds of trade-offs because it lets you think of each ethical dimension as a variable that has its own weight and level of adherence. Because of this, rules can be made that take into account different levels of importance for ethical factors, and trust scores can be changed to reflect this without forcing an all-or-nothing approach. Fuzzy logic is a flexible, rule-based framework that understands how hard it is for AI to make moral decisions, where some principles may be more important in some situations than others.

Additionally, fuzzy logic’s ability to represent doubt and ambiguity fits well with the way ethical problems arise in AI, where information is often lacking and social norms are always changing. AI works in many different settings, each with its own set of ethical rules that can change quickly as social norms and laws change. Because fuzzy logic can deal with uncertainty, the model can adapt to these changing standards and stay useful even as moral priorities change. This ability to change is very important in ethical AI because a model that cannot change could quickly become useless as new ethical issues come up. The trust model is flexible and can reflect the complexity of ethics in the real world because it uses fuzzy logic. This makes it a powerful tool for helping AI systems make decisions.

People trust AI systems more or less based on how fair, private, and open they are. The Fuzzy Trust Model and Ethical Decision Scores framework can be used to look into these issues in more depth. It is seen as a spectrum with different moral levels. An essential component of trust-building is prioritizing privacy, which the model was able to do according to its weights. These moral factors are used by the fuzzy rule-based system in this framework to come up with an ethical decision score. This score can be used to judge the morality of an AI system.

Table 3 shows different situations with different levels of fairness, privacy, and openness, along with the fuzzy trust model scores that go with them. The trust score T(x) was created using a sigmoid membership function, which allowed the trust to grow slowly as the AI got better at being ethical (as shown by the following privacy, fairness, and openness rules). The formula for this membership function was:

Once specific moral requirements are satisfied, there are significant obstacles to a person’s trustworthiness, as this function shows. When an AI system consistently acts ethically, trust goes up sharply. Users will only trust this setup more if it is clear that it follows moral rules. When scores for both privacy and transparency were moderate to high, scores for trust were always higher. These two things have an impact on trust. In real life, this works well for apps in healthcare, finance, and public service where user trust is important.

The ethical decision score E(x) was calculated using the weighted sum of privacy (P), fairness (F), and transparency (Tr). These dimensions were weighed as follows:

w₁ = 0.4 for privacy,

w₂ = 0.3 for fairness, and

w₃ = 0.3 for transparency.

Table 4 shows different mixes of privacy, fairness, and transparency, along with the scores that were used to make ethical decisions. The ethical decision score E(x) was found by adding up the privacy (P), fairness (F), and transparency (Tr) dimensions. After that, we could look at how these moral issues affect AI’s choice-making in general. This model gave privacy the most weight (0.4), which shows its importance for figuring out what is right and wrong. The weights for fairness and transparency were both 0.3 when ethics were taken into account.

The sensitivity analysis demonstrates how variations in privacy adherence significantly affect the trust score T(x) within the AI system. In this analysis, privacy (P) was systematically reduced while keeping fairness (F) and transparency (Tr) constant at 0.6 and 0.8, respectively. By isolating privacy, we aimed to assess its impact on trust, providing insights into how crucial privacy adherence is in maintaining a high trust score, as presented in Table 5.

It is possible to test AI systems on three different moral dimensions: fairness, privacy, and openness. These can be done by using fuzzy rules to make moral decisions. The ethical decision score E(x) changes depending on how well the rules are followed. This is because each rule looks at a different mix of these factors. We can see how these moral issues affect the score as a whole by breaking down each rule (Table 6).

This rule emphasizes that an AI system makes the most ethical choices when it cares a lot about privacy, fairness, and openness. Privacy laws like the GDPR and other similar data governance policies are very strict in places like Germany and Canada.⁴⁵ It is required that systems responsibly collect data, treat users fairly, and be clear about how they work.

Rule 2 says that the ethical decision score goes down if either fairness or privacy is not respected. This indicates a moderate or low outcome. It is clear from this rule that trust cannot exist if there is even one moral aspect that is missing, especially when it comes to privacy or fairness. Data privacy laws may not be very strict in India or some other parts of Latin America.⁴⁶ This means that AI systems may not have strong privacy protections in these places. A system might get a moderate or low ethical score if it does not protect user data well. Users may be given some openness, but when there are not strong privacy and fairness standards in place, they feel vulnerable.

It states that a lack of transparency lowers the ethical decision score directly. Transparency is important because it makes it clear how the system works and how decisions are made. For healthcare AI systems to work in places with strict privacy laws, like the US under Health Insurance Portability and Accountability Act, users must be able to see how decisions are made to keep their trust.^47,48 If a healthcare AI model makes accurate diagnoses but does not explain how it comes to its conclusions in a way that is easy to understand, it may still have trouble earning full trust. Transparency is important for both patients and doctors to make sure that decisions are made correctly and fairly. So, even if privacy and fairness standards are high, trust will stay low if there is not much openness.

The fuzzy logic model did a good job of showing trust as a continuum, with trust scores going up or down depending on how well people followed ethical standards in privacy, fairness, and openness. The results support H1 by showing that trust changes over time as people follow ethical rules more closely. Fuzzy logic is used by the model to cover this range. This is similar to how AI works in the real world, where small changes in moral behaviour lead to higher trust scores. For instance, healthcare systems do not have to do everything at once to build trust; they can take small steps towards being more open over time.

Higher levels of privacy, fairness, and openness always lead to higher trust scores in the model, which supports H2. The fact that ethical adherence and trust are linked in a good way shows that people trust AI systems more when they follow high standards in these areas. For example, people are more likely to trust AI used in financial services if privacy protections are clear, decisions are made fairly, and the loan approval process is kept open.

The sensitivity analysis showed that respecting privacy had the biggest impact on trust, especially when the levels of fairness and transparency were average. The results show that privacy is a very important part of building and keeping trust, especially in areas like healthcare and finance where data is sensitive. A drop in privacy adherence leads to a big drop in trust scores, even when fairness and transparency are moderately upheld. Countries with strict data protection laws, like the European Union’s GDPR, put privacy first in AI applications because they know that users will lose trust if privacy standards are broken, even if other ethical promises are made.⁴⁹ This result backs up H3, showing that privacy is important for keeping people’s trust in AI and that industries that deal with private data should put it ahead of other factors to keep trust.

Finally, the fuzzy rule-based system and sensitivity analysis show that trust in AI systems is complicated and depends on how open, fair, and private they are. The model pretty much backs up the study’s hypotheses because it breaks down the process of making an ethical choice into tests based on rules. The most important thing in trust dynamics was found to be privacy. For this reason, privacy should be the most important thing for data governance frameworks. The trust continuum model in fuzzy logic also shows that as people become more moral, trust grows over time. This supports situations in the real world where making things more moral can slowly boost user confidence. Like what has been seen in places with strict privacy laws, these results make sense. They also show other places how to use good governance to make people trust AI systems more.

While the study does have a few shortcomings, the fuzzy logic framework does a decent job of illustrating the temporal dynamics of trust. One thing to keep in mind is that the importance you place on ethical dimensions can vary from person to person and from one situation to another. To discover the optimal compromise among transparency, equity, and privacy in practice, additional study and feedback from numerous stakeholder groups may be necessary. Potentially, in the future, scientists may investigate machine learning methods that can dynamically adjust these weights in response to trust and real-world data. Second, the fuzzy membership functions used to model trust and morality are based on set limits, in the sigmoid function. Some AI systems might not respond the same way to these limits. For some uses, these thresholds could be even better, but the fuzzy trust model needs more research to be useful in more areas.

One thing that could be done is to see how machine learning and fuzzy logic can be put together to make flexible trust models. Users, regulators, and other interested parties could give real-time feedback to the fuzzy aggregation model. This feedback could be used to change how much weight is given to different ethical dimensions. This is one way that AI systems could keep getting better at being moral and honest as times and standards change. There are other places where the fuzzy trust model could be useful, like schools, healthcare, and finance. It could help us learn more about how trust works in AI systems that are made for certain tasks.

Making moral decisions while online is challenging and fraught with uncertainty; this model illustrates this by using a trust scale and the fact that people’s morals vary. When fuzzy rules are used, AI can tell what is right or wrong based on the situation. In real life, this makes us trust people more. There are now ways to quantify trust, which adds to what is known about AI ethics and data governance. To further enhance the reliability of AI systems in the digital economy, researchers will now investigate ways to incorporate machine learning, specific AI applications, and ethical considerations.