Man-machine trust modeling method based on structured measurement

Abstract

The invention discloses a man-machine trust modeling method based on structured measurement, and relates to the technical field of man-machine interaction. According to the technical scheme, the method comprises the following steps: constructing a structured measurement model containing perception reliability, perception capability, perception scene risk and perception man-machine relationship, quantitatively expressing the man-machine trust level through a mathematical relationship, and dynamically adjusting model parameters according to real-time data, the dynamic influence of multiple factors on trust can be comprehensively considered through the structured trust measurement model, and the method is particularly suitable for complex scenes such as automatic driving; by introducing situation regulation factors (such as risk and social attributes), the adaptability and accuracy of trust evaluation are enhanced; a nonlinear model (Log function) is adopted, so that the explanatory force and the prediction accuracy of the model are improved; a two-way trust concept is introduced, so that man-vehicle trust measurement is closer to practical application, man-machine cooperation and decision intervention can be better supported, and then the evaluation precision and the application range of man-machine trust are improved.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, and in particular to a human-computer trust modeling method based on structured measurement. Background Technology

[0002] Trust, as a core factor in human-machine collaboration, is influenced by multiple dimensions, encompassing individual characteristics, technological attributes, contextual features, and the human-machine relationship. However, the lack of a clear and systematic structured measurement framework in current technologies makes it difficult to effectively integrate and dynamically track these influencing factors, hindering their support for subsequent predictive modeling and intervention design. Therefore, a structured measurement-based trust modeling method is urgently needed to quantify trust and support technological optimization and decision-making interventions in related fields.

[0003] Currently, structured trust measurement models have become one of the key methods in trust research, with the McKinsey Trust Equation widely regarded as a classic representative of structured trust models. This model defines trust through the ratios of factors such as credibility, reliability, intimacy, and self-direction, and has been widely applied in practical scenarios such as organizational management and customer service. However, this model fails to effectively consider the impact of situational factors (such as environmental risks and social interactions) on trust, limiting its applicability in complex human-computer interaction scenarios. Summary of the Invention

[0004] The purpose of this invention is to provide a human-machine trust modeling method based on structured measurement. By introducing multi-dimensional factors (including perceived reliability, perceived capability, human-machine relationship, scenario risk, and social attributes), a predictive model capable of dynamically tracking and accurately quantifying trust levels is established. This invention is particularly applicable to high-risk scenarios with close human-machine interaction, such as autonomous driving, and can effectively improve the accuracy and applicability of human-machine trust assessment.

[0005] This invention provides a human-computer trust modeling method based on structured measurement, comprising the following steps:

[0006] A structured measurement model for human-machine trust is constructed, which quantifies human-machine trust as a function of at least four core variables. These core variables include: perceived reliability, representing the user's subjective assessment of the consistency and reliability of the trusted party's behavior; perceived capability, representing the user's subjective assessment of the trusted party's professional knowledge, skills, and task ability; perceived scenario risk, representing the user's subjective assessment of the potential risks borne due to inconsistencies in external environmental conditions, incomplete information, and unpredictability of behavioral outcomes; and perceived human-machine relationship, representing the user's subjective assessment of whether the trusted party prioritizes its own interests or the user's interests. The quantitative model of human-machine trust is expressed as: Trust = (Perceived Reliability + Perceived Capability) / (Perceived Scenario Risk + Perceived Human-Machine Relationship).

[0007] Preferably, perceived reliability corresponds to the system reliability of the trusted party, that is, the ability of the trusted party to maintain stable performance and be fault-free in repetitive tasks.

[0008] Preferably, the perception capability is related to the system capability of the trusted party, that is, whether the trusted party possesses the knowledge, strategies, and execution capabilities required to complete the task.

[0009] Preferably, the perception of human-machine relationship is reflected through control led by the trusted party or the user.

[0010] Preferably, it also includes the introduction of a perceived scenario risk attribute as a modifier, wherein the perceived scenario risk attribute represents the subjective assessment of the risks borne in a specific task due to external environmental uncertainty, information incompleteness, and unpredictability of behavioral outcomes.

[0011] Preferably, after introducing the perception of the other party's trust, the quantitative model of human-machine trust is: Trust = (perceived reliability + perceived capability + perceived other party's trust) / (perceived scenario risk + perceived human-machine relationship), where perceived other party's trust represents the trusted party's perception of the trusting party's trust in the user, and is used to realize two-way trust modeling.

[0012] Preferably, the model further includes performing a logarithmic transformation on at least one variable in the quantization model to form a nonlinear model.

[0013] Preferably, the method further includes verifying the fitting effect of the quantization model, which includes the following steps: acquiring experimental data; calculating a confidence prediction value based on the quantization model; comparing the confidence prediction value with the actual measured confidence level, and using at least one of the following: correlation coefficient, coefficient of determination, mean square, or root mean square error to evaluate the accuracy of the model.

[0014] Compared with the prior art, the beneficial effects of this application are as follows:

[0015] Systematic integration of multi-dimensional variables: By quantifying trust as a function of core variables, the problem of scattered factors and difficulty in integration in existing technologies is solved, providing a unified framework for trust assessment.

[0016] Dynamically adapting to complex scenarios: By introducing the perceived risk of the scenario as a moderating factor and supporting nonlinear model transformation, the adaptability and evaluation accuracy of the model in dynamic environments are improved.

[0017] Achieving two-way trust modeling: By introducing the perception of the other party's trust, the model can simultaneously reflect the trust relationship between the user and the trusted party, as well as the trusted party and the user, thus enhancing the comprehensiveness of the model.

[0018] The verification method is reliable: Through regression fitting and error evaluation, the scientific validity and practicality of the model are ensured, providing data support for practical applications.

[0019] In summary, this invention effectively solves the problem of the lack of a systematic measurement framework in the prior art, which makes it difficult to integrate and dynamically track influencing factors, and significantly improves the applicability and evaluation accuracy of the model in complex human-computer interaction scenarios.

[0020] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0023] Figure 1 This is a measurement effect diagram of the human-machine trust structured measurement model in an embodiment of the present invention. Detailed Implementation

[0024] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0025] Example 1:

[0026] This embodiment provides a human-machine trust modeling method based on structured measurement. This method quantifies human-machine trust into a function of four core variables by constructing a structured measurement model of human-machine trust. The specific steps are as follows:

[0027] Building a structured measurement model

[0028] The model represents human-machine trust as a function of the following four core variables:

[0029] Perceived reliability: Characterizes a user's subjective assessment of the consistency and reliability of the behavior of a trusted party. For example, in autonomous driving scenarios, users assess reliability by observing whether the vehicle maintains stable acceleration, braking, and steering behavior under the same road conditions.

[0030] Perceived capability: This represents a user's subjective assessment of the trusted party's expertise, skills, and task capabilities. For example, users judge a system's capability based on whether the vehicle accurately identifies traffic signs and avoids obstacles.

[0031] Perceived scenario risk: This refers to the user's subjective assessment of the potential risks they face due to external environmental factors (such as severe weather or complex road conditions). For example, a user might perceive the risk of using an autonomous driving system to be higher in heavy rain.

[0032] Perceived human-machine relationship: This represents a user's subjective assessment of whether the trusted party prioritizes their own interests or the user's interests. For example, if a vehicle prioritizes passenger safety over traffic efficiency, the user may perceive that the vehicle prioritizes the user's interests.

[0033] Quantization model

[0034] The model of human-machine trust is represented as follows:

[0035] Trust = (Perceived reliability + Perceived capability) / (Perceived scenario risk + Perceived human-machine relationship).

[0036] Each variable is scored using a Likert scale (1-5 points), for example:

[0037] Perceived reliability: 1 point (completely unreliable) to 5 points (completely reliable);

[0038] Perception ability: 1 point (completely lacking ability) to 5 points (completely possessing ability);

[0039] Perceived risk level: 1 point (no risk) to 5 points (extremely high risk);

[0040] Perception of human-computer relationship: 1 point (completely focused on self-interest) to 5 points (completely focused on user interests).

[0041] Implementation effect

[0042] This model allows users to dynamically assess their level of trust in human-machine systems, for example:

[0043] When perceived reliability = 4, perceived capability = 4, perceived scenario risk = 2, and perceived human-machine relationship = 5, the trust value = (4+4) / (2+5) = 1.14;

[0044] When the perceived risk level of the scenario increases to 4, the trust value = (4+4) / (4+5) = 0.89, indicating that the increase in risk level will decrease the trust value.

[0045] Example 2: Correspondence between perceived reliability and system reliability

[0046] In this embodiment, perceived reliability further corresponds to the system reliability of the trusted party, that is, the ability of the trusted party to maintain stable performance and be fault-free in repetitive tasks. For example:

[0047] Autonomous driving system: By recording the vehicle's braking response time in 100 instances under the same road conditions (e.g., the target is 2 ± 0.5 seconds), if more than 90% of the response times fall within the range, the system has high reliability, and the user's perceived reliability score can reach 4-5 points.

[0048] Industrial robots: The positioning error of the robotic arm in repetitive assembly tasks (e.g., target error ≤ 0.1mm) is considered to be high by the user if the error fluctuation range is small.

[0049] Example 3: Correspondence between perception ability and system capability

[0050] In this embodiment, perception capability corresponds to the system capability of the trusted party, that is, whether the trusted party possesses the knowledge, strategies, and execution capabilities required to complete the task. For example:

[0051] Autonomous driving systems: Testing whether the vehicle can complete tasks in the following scenarios:

[0052] Complex road conditions (such as intersections without traffic lights);

[0053] Special missions (such as emergency avoidance, automatic parking).

[0054] If the task success rate reaches 95% or higher, the user's perception score can reach 4-5 points.

[0055] Medical assistance system: The user perception score is high by evaluating the accuracy of the AI ​​diagnostic model in 1,000 cases (e.g., ≥90%).

[0056] Example 4: Implementation of Interactive Control Based on Human-Machine Relationship Perception

[0057] In this embodiment, the perception of human-computer relationship is manifested by controlling the interaction method led by the trusted party or the user. For example:

[0058] Trust-driven: The autonomous driving system automatically adjusts the vehicle speed and route, and the user only provides the destination. In this case, the user may think that the system is more concerned with its own efficiency (such as reducing travel time), and the perceived human-machine relationship score is low (such as 1-3 points).

[0059] User-centric: The system provides multiple route options for users to choose from, and users can take over control at any time. At this time, users may feel that the system pays more attention to their needs, and their perceived human-computer relationship score is higher (e.g., 4-5 points).

[0060] Example 5: Introducing perceived scenario risk as a moderating factor

[0061] In this embodiment, the perceived risk of the scenario is also introduced as a moderating factor to characterize the degree of risk in the task environment and scenario. For example:

[0062] High-risk scenarios: When autonomous vehicles are in operation, the road traffic environment is complex, there are many vehicles on the road, and congestion and emergencies may occur. In this case, the perception scenario risk attribute is high (e.g., 4-5 points), which may reduce trust in the autonomous driving system.

[0063] Low-risk scenarios: Good road conditions, straight road surface, and almost no other vehicles driving. The perceived risk of the scenario is low (e.g., 1-2 points), and the trust value may be higher.

[0064] Example 6: Two-way Trust Modeling

[0065] In this embodiment, the concept of perceived trust (the trusted party's trust in the user) is introduced to achieve two-way trust modeling, and the quantification model is as follows:

[0066] Trust = (Perceived reliability + Perceived capability + Perceived trust in the other party) / (Perceived risk of the scenario + Perceived human-machine relationship).

[0067] For example:

[0068] Perceived trust in the other party: This is assessed by whether the autonomous driving system allows users to take over control at any time. If the system frequently prompts the user to take over, the perceived trust in the other party is low (e.g., 1-3 points).

[0069] Application scenario: In emergency medical rescue, the system needs to trust the user's instructions (such as detouring through congested areas), and in this case, the perceived trust score of the other party is relatively high (such as 4-5 points).

[0070] Example 7: Nonlinear Model Optimization

[0071] In this embodiment, the variables in the quantization model are logarithmically transformed to form a nonlinear model. For example:

[0072] Conversion formula: Trust = log(1 + (perceived reliability + perceived capability) / (perceived scenario risk + perceived human-machine relationship)).

[0073] Effect: When the scene and system stimulation are high, the marginal effect of users' psychological expectations decreases. Logarithmic transformation can reflect the mathematical formula of this diminishing marginal benefit, which is more in line with users' psychology.

[0074] Example 8: Model Accuracy Verification

[0075] In this embodiment, the fitting effect of the quantization model is tested, and the steps are as follows:

[0076] Obtain experimental data: Recruit 100 users and collect their trust ratings (actual values) of the human-computer system and ratings of four core variables through a questionnaire survey.

[0077] Calculate the predicted value: Calculate the predicted trust value for each user based on the quantification model.

[0078] Evaluation accuracy: using the coefficient of determination (R²) 2 ) Evaluate the model fit; if Rfit 2 A value of ≥0.8 indicates that the model has high accuracy.

[0079] Example 9: The trusted party is an autonomous driving system

[0080] In this embodiment, the trusted party is the autonomous driving system, and specific application scenarios include:

[0081] Highway autonomous driving: Users assess reliability by perceiving whether the vehicle maintains lane stability and automatically follows other vehicles;

[0082] Autonomous driving on urban roads: Users assess the system's capabilities by evaluating whether the vehicle can recognize pedestrians, non-motorized vehicles, etc.

[0083] Extreme weather scenarios: Users assess the risk level by observing the system's performance in heavy rain and fog.

[0084] Through the above embodiments, the present invention provides a quantifiable and interpretable human-machine trust modeling method, which is applicable to various scenarios such as autonomous driving, medical assistance, and industrial control.

[0085] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A human-computer trust modeling method based on structured measurement, characterized in that, Includes the following steps: A structured measurement model for human-machine trust is constructed, which quantifies human-machine trust as a function of at least four core variables, including: perceived reliability, which represents the user's subjective assessment of the consistency and reliability of the trusted party's behavior; Perceived ability represents the user's subjective assessment of the trusted party's professional knowledge, skills, and task capabilities; perceived scenario risk represents the user's subjective assessment of the potential risks borne by the user due to incomplete external environmental conditions, incomplete information, and unpredictable behavioral outcomes; perceived human-machine relationship represents the user's subjective assessment of whether the trusted party is more concerned with its own interests or the user's interests; the quantitative model of human-machine trust is expressed as: Trust = (Perceived reliability + Perceived ability) / (Perceived scenario risk + Perceived human-machine relationship).

2. The human-machine trust modeling method based on structured measurement according to claim 1, characterized in that, Perceived reliability corresponds to the system reliability of the trusted party, that is, the ability of the trusted party to maintain stable performance and be fault-free in repetitive tasks.

3. The method according to claim 1, characterized in that, The perception capability refers to the system capability of the trusted party, that is, whether the trusted party possesses the knowledge, strategies, and execution capabilities required to complete the task.

4. The human-machine trust modeling method based on structured measurement according to claim 1, characterized in that, The perception of human-machine relationship is manifested through control that is led by the trusted party or the user.

5. The human-machine trust modeling method based on structured measurement according to claim 1, characterized in that, It also includes the introduction of perceived scenario risk attributes as a modifier, which represent the subjective assessment of the risks borne in a specific task due to external environmental uncertainty, information incompleteness, and unpredictability of behavioral outcomes.

6. The human-machine trust modeling method based on structured measurement as described in claim 5, characterized in that, After introducing the concept of perceived trust in the other party, the quantitative model of human-machine trust is: Trust = (Perceived reliability + Perceived capability + Perceived trust in the other party) / (Perceived scenario risk + Perceived human-machine relationship), where perceived trust in the other party represents the trusted party's perception of the trusting party's trust in the user, and is used to realize two-way trust modeling.

7. The human-machine trust modeling method based on structured measurement according to claim 1, characterized in that, It also includes performing a logarithmic transformation on at least one variable in the quantization model to form a nonlinear model.

8. The human-machine trust modeling method based on structured measurement according to claim 1, characterized in that, It also includes verifying the fitting effect of the quantization model, which includes the following steps: acquiring experimental data; calculating the confidence prediction value based on the quantization model; comparing the confidence prediction value with the actual measured confidence level, and using at least one of the correlation coefficient, determination coefficient, mean square or root mean square error to evaluate the accuracy of the model.

Images