Trust is confidence enough to proceed.
We rarely have enough time, information or access to verify everything for ourselves. We depend upon other people, institutions, instruments and systems. We accept information, follow advice, delegate decisions and act on representations that somebody else has helped to construct. Trust makes this possible. To trust is not necessarily to believe that someone is infallible or completely honest. It is to judge that relying upon them is acceptable for the purpose and stakes involved. In this sense, trust is a commitment made under uncertainty.Every communication has two subjects
Every communication has two subjects: the subject of the message and the sender of the message. A receiver may evaluate both: • Is this claim plausible? • Is this person credible? • What might they know? • What might they want? • Does the message fit with other available evidence? • Does the way it has been communicated appear normal? The receiver constructs a representation not only of what is being said, but also of who is saying it and why. Trust in the message and trust in the sender may support one another—but they are not the same thing. A trusted person can be mistaken. An untrusted person can tell the truth.Trust conserves finite resources
Checking everything would consume more time and attention than we possess. Trust allows us to stop investigating and proceed. We trust that a familiar road remains open, that a colleague has completed an agreed task, that a measurement is accurate or that a source has represented the evidence fairly. This creates efficiency, but also vulnerability. The same mechanism that allows cooperation can be exploited by someone who understands what the receiver expects a trustworthy person, message or situation to look like.Deception is designed for a receiver
A deception must usually pass through another person’s model of normal. The sender may imitate familiar language, behaviour, credentials, documents, relationships or technical signals. They may provide enough accurate information to make the false part appear credible. They may create urgency so that the receiver commits before investigating. Successful deception does not need to reproduce reality perfectly. It needs to produce a representation that is good enough for the intended receiver to accept. Different receivers may therefore respond differently to the same attempt. What appears convincing to one person may contain an obvious departure from normal for another.Recognition begins with difference
Whether we are trying to identify something as false or verify it as genuine, the cognitive task has a common structure: detecting meaningful departures from a model of normal. An inconsistency may snag us: an unexpected request, a strange tone, an unfamiliar payment method, an incorrect detail or behaviour that does not fit the person we thought we knew. But difference alone does not prove deception. Honest people make mistakes. Unusual events occur. Genuine messages sometimes look suspicious, and sophisticated deception may look entirely ordinary. The receiver must still judge what the difference means and whether further investigation is worth its cost.Fake, real and authentic
How do we decide whether something is what it appears to be? “Fake” and “real” can sound like simple opposites, but authenticity is rarely a single binary judgment. Something may be: • created by a real person but contain false information; • an authentic photograph presented in a false context; • an accurate copy rather than the original; • partly altered while retaining genuine elements; • factually mistaken without being deliberately deceptive; • artificially generated but clearly labelled; • genuine in origin but used to create a misleading impression. We may therefore need to ask several different questions: • Did this object, message or recording come from the claimed source? • Has it been altered? • Is its content accurate? • Is important context missing? • Is it being presented for a deceptive purpose? • Is it the original, a copy or a reconstruction? • What exactly are we calling real? Apparent normality cannot settle these questions. A sophisticated fake may reproduce all the visible or audible features the receiver expects. At the same time, an unfamiliar but genuine item may be rejected because it does not fit the receiver’s model of normal. The problem is therefore not merely detecting difference. It is deciding which differences carry useful information—and recognising when the absence of an obvious difference tells us very little.Provenance
As documents, images, voices and messages become easier to reproduce, familiar appearance may become a weaker signal of authenticity. The question may shift from: Does this look or sound real? to: What evidence shows where this came from and what happened to it along the way? Provenance concerns origin and history. It may include who created something, how it was recorded, where it has been stored, whether it has been altered and how it reached the receiver. Provenance signals can become part of a model of normal. But those signals must themselves be understood and trusted. They may be incomplete, misinterpreted or imitated. No signal verifies itself. Artificial intelligence makes this problem more urgent because it can generate signals that closely match familiar models of normal. But the underlying challenge is older and broader than AI. Forgery, impersonation, counterfeit goods, misleading context and manufactured evidence have always depended upon what a receiver is prepared to accept as genuine.Trust and deception through the NMF
Detection What appeared normal, familiar or credible—and what, if anything, differed? A deception may succeed because no difference is detected. Trust may also fail because a harmless difference is treated as evidence of dishonesty. Evaluation What might the signal mean? The receiver may consider error, misunderstanding, coincidence, incompetence, manipulation or deliberate deception. Their interpretation will depend upon their existing model of the sender and situation. Commitment Is there enough confidence to trust, challenge, verify, refuse or wait? The required threshold may change with the stakes. Accepting a restaurant recommendation does not require the same confidence as transferring money or authorising access to a secure system. Action The receiver may proceed, seek confirmation, ask another person, test the claim, restrict access, refuse the request or deliberately accept the remaining uncertainty. Each action consumes resources and creates different risks. Learning How does the outcome alter future trust? A successful interaction may strengthen trust. A betrayal may damage it. But outcomes can teach the wrong lesson. A deception that remains undiscovered may reinforce trust in the deceiver. A coincidence may make an innocent person appear dishonest. Repeated success may create overconfidence and lower future scrutiny. Trust today and trust tomorrow A sender may be able to gain compliance today by exaggerating, concealing or manipulating. But every interaction can also alter the receiver’s future model of the sender. Communication therefore has at least two possible objectives: • securing the desired response now; and • preserving sufficient trust for future interaction. A message that succeeds today may make tomorrow’s message less credible. Conversely, admitting uncertainty or delivering unwelcome information may reduce immediate influence while strengthening longer-term trust. Authenticity and provenance As messages, images, voices and documents become easier to reproduce, familiarity alone may become a weaker signal of authenticity. The question may shift from: Does this look or sound genuine? to: What reliable evidence shows where this came from and what happened to it along the way? Provenance signals can become part of the receiver’s model of normal. But these signals must themselves be understood, trusted and protected from imitation. No signal verifies itself.Questions NMF can bring to trust and deception
1. What does the receiver expect a trustworthy source to look like? 2. Which signals are being used as evidence of credibility? 3. Are those signals actually connected to trustworthiness? 4. What is known about the message? 5. What is merely inferred about the sender? 6. What does the sender want the receiver to notice? 7. What might the sender prefer the receiver not to notice? 8. What resources would verification require? 9. What are the consequences of trusting incorrectly? 10. What are the consequences of distrusting incorrectly? 11. What confidence threshold is appropriate for the stakes? 12. Is urgency preventing further investigation? 13. How will this interaction affect trust tomorrow? 14. What will the outcome genuinely justify learning?A potential use — not a lie detector
The Normal-Model Framework cannot determine whether a person is honest, whether a message is true or whether a source deserves trust. It does not provide a reliable list of behavioural signs that reveal deception. Nor does it replace evidence, verification, professional investigation or technical security controls. It offers a way to examine how credibility is constructed, how trust crosses a threshold, how deception may exploit models of normal and how present interactions shape future judgment. Its potential contribution lies in making the commitment visible: What am I being asked to accept? Why does it appear credible? What would justify checking? And what am I prepared to place at risk by trusting—or refusing to trust?
Trust &
Deception
“Trust lets us act without checking everything. That is what
makes it useful - and also what makes us vulnerable.”
“Something can be real and still
be used to mislead you.”
© Normal-Model Framework - 2026
Trust is confidence enough to
proceed.
We rarely have enough time, information or access to verify everything for ourselves. We depend upon other people, institutions, instruments and systems. We accept information, follow advice, delegate decisions and act on representations that somebody else has helped to construct. Trust makes this possible. To trust is not necessarily to believe that someone is infallible or completely honest. It is to judge that relying upon them is acceptable for the purpose and stakes involved. In this sense, trust is a commitment made under uncertainty.Every communication has two
subjects
Every communication has two subjects: the subject of the message and the sender of the message. A receiver may evaluate both: • Is this claim plausible? • Is this person credible? • What might they know? • What might they want? • Does the message fit with other available evidence? • Does the way it has been communicated appear normal? The receiver constructs a representation not only of what is being said, but also of who is saying it and why. Trust in the message and trust in the sender may support one another—but they are not the same thing. A trusted person can be mistaken. An untrusted person can tell the truth.Trust conserves finite resources
Checking everything would consume more time and attention than we possess. Trust allows us to stop investigating and proceed. We trust that a familiar road remains open, that a colleague has completed an agreed task, that a measurement is accurate or that a source has represented the evidence fairly. This creates efficiency, but also vulnerability. The same mechanism that allows cooperation can be exploited by someone who understands what the receiver expects a trustworthy person, message or situation to look like.Deception is designed for a receiver
A deception must usually pass through another person’s model of normal. The sender may imitate familiar language, behaviour, credentials, documents, relationships or technical signals. They may provide enough accurate information to make the false part appear credible. They may create urgency so that the receiver commits before investigating. Successful deception does not need to reproduce reality perfectly. It needs to produce a representation that is good enough for the intended receiver to accept. Different receivers may therefore respond differently to the same attempt. What appears convincing to one person may contain an obvious departure from normal for another.Recognition begins with difference
Whether we are trying to identify something as false or verify it as genuine, the cognitive task has a common structure: detecting meaningful departures from a model of normal. An inconsistency may snag us: an unexpected request, a strange tone, an unfamiliar payment method, an incorrect detail or behaviour that does not fit the person we thought we knew. But difference alone does not prove deception. Honest people make mistakes. Unusual events occur. Genuine messages sometimes look suspicious, and sophisticated deception may look entirely ordinary. The receiver must still judge what the difference means and whether further investigation is worth its cost.Fake, real and authentic
How do we decide whether something is what it appears to be? “Fake” and “real” can sound like simple opposites, but authenticity is rarely a single binary judgment. Something may be: • created by a real person but contain false information; • an authentic photograph presented in a false context; • an accurate copy rather than the original; • partly altered while retaining genuine elements; • factually mistaken without being deliberately deceptive; • artificially generated but clearly labelled; • genuine in origin but used to create a misleading impression. We may therefore need to ask several different questions: • Did this object, message or recording come from the claimed source? • Has it been altered? • Is its content accurate? • Is important context missing? • Is it being presented for a deceptive purpose? • Is it the original, a copy or a reconstruction? • What exactly are we calling real? Apparent normality cannot settle these questions. A sophisticated fake may reproduce all the visible or audible features the receiver expects. At the same time, an unfamiliar but genuine item may be rejected because it does not fit the receiver’s model of normal. The problem is therefore not merely detecting difference. It is deciding which differences carry useful information—and recognising when the absence of an obvious difference tells us very little.Provenance
As documents, images, voices and messages become easier to reproduce, familiar appearance may become a weaker signal of authenticity. The question may shift from: Does this look or sound real? to: What evidence shows where this came from and what happened to it along the way? Provenance concerns origin and history. It may include who created something, how it was recorded, where it has been stored, whether it has been altered and how it reached the receiver. Provenance signals can become part of a model of normal. But those signals must themselves be understood and trusted. They may be incomplete, misinterpreted or imitated. No signal verifies itself. Artificial intelligence makes this problem more urgent because it can generate signals that closely match familiar models of normal. But the underlying challenge is older and broader than AI. Forgery, impersonation, counterfeit goods, misleading context and manufactured evidence have always depended upon what a receiver is prepared to accept as genuine.Trust and deception through the
NMF
Detection What appeared normal, familiar or credible—and what, if anything, differed? A deception may succeed because no difference is detected. Trust may also fail because a harmless difference is treated as evidence of dishonesty. Evaluation What might the signal mean? The receiver may consider error, misunderstanding, coincidence, incompetence, manipulation or deliberate deception. Their interpretation will depend upon their existing model of the sender and situation. Commitment Is there enough confidence to trust, challenge, verify, refuse or wait? The required threshold may change with the stakes. Accepting a restaurant recommendation does not require the same confidence as transferring money or authorising access to a secure system. Action The receiver may proceed, seek confirmation, ask another person, test the claim, restrict access, refuse the request or deliberately accept the remaining uncertainty. Each action consumes resources and creates different risks. Learning How does the outcome alter future trust? A successful interaction may strengthen trust. A betrayal may damage it. But outcomes can teach the wrong lesson. A deception that remains undiscovered may reinforce trust in the deceiver. A coincidence may make an innocent person appear dishonest. Repeated success may create overconfidence and lower future scrutiny. Trust today and trust tomorrow A sender may be able to gain compliance today by exaggerating, concealing or manipulating. But every interaction can also alter the receiver’s future model of the sender. Communication therefore has at least two possible objectives: • securing the desired response now; and • preserving sufficient trust for future interaction. A message that succeeds today may make tomorrow’s message less credible. Conversely, admitting uncertainty or delivering unwelcome information may reduce immediate influence while strengthening longer-term trust. Authenticity and provenance As messages, images, voices and documents become easier to reproduce, familiarity alone may become a weaker signal of authenticity. The question may shift from: Does this look or sound genuine? to: What reliable evidence shows where this came from and what happened to it along the way? Provenance signals can become part of the receiver’s model of normal. But these signals must themselves be understood, trusted and protected from imitation. No signal verifies itself.Questions NMF can bring to trust
and deception
1. What does the receiver expect a trustworthy source to look like? 2. Which signals are being used as evidence of credibility? 3. Are those signals actually connected to trustworthiness? 4. What is known about the message? 5. What is merely inferred about the sender? 6. What does the sender want the receiver to notice? 7. What might the sender prefer the receiver not to notice? 8. What resources would verification require? 9. What are the consequences of trusting incorrectly? 10. What are the consequences of distrusting incorrectly? 11. What confidence threshold is appropriate for the stakes? 12. Is urgency preventing further investigation? 13. How will this interaction affect trust tomorrow? 14. What will the outcome genuinely justify learning?A potential use — not a lie detector
The Normal-Model Framework cannot determine whether a person is honest, whether a message is true or whether a source deserves trust. It does not provide a reliable list of behavioural signs that reveal deception. Nor does it replace evidence, verification, professional investigation or technical security controls. It offers a way to examine how credibility is constructed, how trust crosses a threshold, how deception may exploit models of normal and how present interactions shape future judgment. Its potential contribution lies in making the commitment visible: What am I being asked to accept? Why does it appear credible? What would justify checking? And what am I prepared to place at risk by trusting—or refusing to trust?
Trust
& Deception
“Trust lets us act without checking everything. That is what
makes it useful - and also what makes us vulnerable.”
“Something can be real and still
be used to mislead you.”
