Exploring the Ethical and Legal Boundaries of Artificial Intelligence in Forensic and Document Examination

Artificial intelligence has become an inevitable trend in modern society. As a technical means, artificial intelligence has been continuously developed and applied. And the application of artificial intelligence in judicial appraisal and document examination is also increasing [1-4].

Judicial appraisal refers to the scientific identification of relevant issues by personnel with specialized knowledge and skills according to the needs of a particular case and the provisions of the judicial process, to provide scientific opinions for the judiciary [5-7]. Document examination is an important means of criminal forensic examination of physical evidence, which includes handwriting test, seal imprint test, printed document test, document formation time test, defacement document test, speech recognition and analysis and other aspects [8-11]. It aims to solve the problem of the authenticity of documents and physical evidence, and provide evidence support for the investigation, prosecution, trial, mediation, etc. of criminal or civil cases [12-13]. In recent years, scientific and technological innovation represented by artificial intelligence technology has developed very rapidly, bringing great changes to people’s work, study and lifestyle [14-15]. As far as the current situation of artificial intelligence technology involved in forensic identification is concerned, it promotes the intelligent transformation of the forensic identification field, improves the efficiency of forensic identification work, enhances the quality of forensic identification work, reduces forensic identification corruption, and promotes the progress of forensic identification science and technology [16-19]. However, AI technology also faces ethical practical dilemmas such as unclear legal status of forensic subject, unstable legal system of forensic appraisal, and legal data security threats [20-22]. Therefore, the legal system system of AI participation in forensic identification should be actively constructed to meet the challenges facing forensic identification in the era of artificial intelligence [23-24]. Specifically, the legal-led judicial appraisal system of artificial intelligence should be established, the attribution of responsibility for artificial intelligence judicial appraisal should be clarified, the information security management mechanism of artificial intelligence should be constructed, and the database of artificial intelligence judicial appraisal should be constructed and so on [25-28].

This paper first clarifies the connotation and extension of digital intelligence appraisal, revealing the double-edged sword effect of digital intelligence appraisal. From the dimension of judicial ethics, it analyzes the improvement elements of the judicial appraisal initiation system and emphasizes the importance of judicial ethics in the digitalization of judicial appraisal. The improved AHP method and fuzzy comprehensive evaluation method are introduced to establish the ethical risk evaluation index system. X case is selected for application analysis to explore its ethical risk. Based on the stakeholder theory, design the ethical risk evaluation indexes of numerical intelligence judicial appraisal. Dynamically identify the key stakeholders in each stage for the determination of ethical risk factors and the division of risk sources. Based on expert scores, subjective-objective combination assignment method is used to assign weights to the indicators. Utilizing intuitionistic fuzzy set theory to calculate the ethical risk value and evaluate the ethical risk of X case. Combined with the application of the ethical risk assessment model of numerical intelligence forensic appraisal, the legal and ethical supervision strategy of artificial intelligence intervention in forensic appraisal is proposed.

2

Transformation of the digital intelligence of forensics in the context of judicial ethics

2.1

Digital intelligence transformation of the core aspects of forensics

The currently established appraisal opinion generation mechanism can better abstract the elements and structure of traditional appraisal. However, with the deep integration of digital intelligence technology into judicial appraisal, the applicability of the generation mechanism has been challenged. First, it is difficult to explain the algorithmic elements that appear in more and more appraisals. Second, it is difficult to reflect the interaction between humans and algorithms, which in turn fails to indicate the hidden reliability risks of digital forensics and provide an analytical framework for reviewing and judging the resulting appraisal opinions. Therefore, it is necessary to follow the trend of digital intelligence technology and update the mechanism of digital intelligence appraisal.

The biggest difference between digital appraisal and traditional appraisal is that algorithms intervene in the test (stage), which leads to the exclusive status of the appraiser to be impacted. When the algorithm intervenes in the judicial appraisal, digital intelligence algorithm technology not only plays the role of auxiliary appraisers, but also substantially involved in the test of the decision-making - its specific embodiment in the following three aspects. 1)

Feature recognition automation

Automation is the use of technology, programs, robotics or processes to automatically identify features, in the case of minimal human input to achieve feature identification. On the one hand, algorithms can automatically recognize human-perceivable features. On the other hand, algorithms can also automate the identification of features that are difficult for humans to perceive directly. During identification, information such as images and speech can be captured directly by human eyes and ears, but dynamic features such as writing speed, pressure and timing hidden in handwriting, and resonance peaks, tone intensity and base frequency information contained in speech are difficult to be perceived directly by human eyes or ears. With the embedding of digital intelligence technology, the above features that could not be or are difficult to be reflected on the object can be recognized and recorded by algorithm-driven sensors, which undoubtedly expands the frontier of features that can be used for comparison and analysis.

2)

Automation of comparison and analysis

The features extracted or recognized by the algorithm will be used for automatic matching or analysis. Automatic comparison, refers to the matching algorithm will be samples features and database archived sample features for automated comparison, screening out the database is most similar to a series of samples to be examined, and will be arranged in accordance with the order of similarity.

Algorithms can be used not only for automated comparison, but also for automated analysis of other specialized problems. It should be noted, however, that both automatic comparison and automatic analysis do not mean that there is no involvement of experts such as expert witnesses in this process. The reliable application of the algorithm, without experts such as experts to set the algorithm-related parameters or assumptions in advance, to confirm that the input samples of data applicable to a particular algorithm. Therefore, the algorithm automatically compared to analyze the “identification results”, is a kind of algorithmic analysis and human knowledge interacting with the conclusion of the opinion.

3)

Authenticator verification

Automatic identification of features and automatic comparison and analysis can be collectively referred to as automated decision-making, it should be noted that, although automated decision-making can be given such as the identification of the same “identification results”, but the identification results do not mean that the identification of the same “identification opinion”. Generally speaking, the algorithm will be based on the input samples data, according to the samples and samples of the degree of similarity or other criteria, to return to a (suspicious person) “candidate list”, the appraiser and then based on the traditional comparative test methods, to determine the samples and samples whether they originate from the same person. For identification beyond the cognitive ability of the appraiser, the appraiser also has to play the role of verification of space: before the algorithm runs, the need for the appraiser to set the relevant parameters or to make specific assumptions, and to ensure that the case of the samples are applicable to the algorithm’s reliability range; in the algorithm is running, the appraiser is also necessary to verify that the algorithm is running normally, in order to issue the relevant appraisal opinion.

Due to the intervention of algorithmic technology, the generation mechanism of digital intelligence judicial appraisal is no longer the “expert global domination” structure, which is significantly different from the traditional appraisal. The mechanism of generating digital-intelligent forensic opinions can be summarized as “automated decision-making + appraiser verification”.

2.2

Forensic initiation system

In the initiation system of digital judicial appraisal, how the exercise of the judge’s decision-making power can bring the legal system, the spirit of the law and the judicial case into unity in fact and value is the key to perfecting China’s initiation system of digital judicial appraisal.

To improve China’s judicial appraisal initiation system, the first step is to make the legal system itself conform to the spirit of the legislation and match the values pursued by judicial ethics. Judicial ethics require judges to maintain neutrality in litigation, and China’s judicial appraisal initiation, the court have the right to initiate the appraisal process without the application of the parties, this initiative behavior makes in the litigation Shen court or the judge neutrality of the shaking. In this case, the question is how to improve the judicial system and make it and the judicial ethical values consistent.

To make the judicial system and ethical values consistent can not simply control the judge’s decision-making power, but should be to make the rights of all parties in the litigation to achieve a balance. In China’s judicial practice, by the civil law system of the court mode of influence, decided that the judge on the judicial appraisal of the initiation of the authority of the mode will not change. In this case need to expand the party’s right to apply, in addition to the statutory matters, the party enjoys the right to apply for judicial appraisal, the judge enjoys the right to decide whether to carry out judicial appraisal, but its decision depends on the party’s application rather than the authority.

The difference in the mode of initiation of judicial appraisal after such reform is that the basis of the judge’s decision-making power has changed, from the competence of the court and the judge to the application of the parties. At the same time, the law for the court’s right to initiate should also be necessary constraints and specific provisions, for the decision not to entrust the identification of the reasons should be given, and to give the parties a way to relief, such as appeals, complaints and so on.

For the management and training of individual judges, it should go beyond the purely technical model of judge management and training. In order for judges to achieve the unity of facts and values in exercising their decision-making power in the judicial appraisal of numerical intelligence, they should be helped to correct the judicial ethical issues embedded in the process of judicial activities, and to understand the humanistic values of justice as determined by judicial ethics, so that they can continue to explore the mode of adjudication and judicial behaviors in line with judicial ethics, thus promoting the improvement and development of the judicial appraisal system.

Improving the conflict between the right to apply and the right to decide in the judicial appraisal initiation system, so as to make it consistent, is an important issue in the reform of China’s judicial system. The solution of this problem requires not only the improvement of the system, but also the intervention of judicial ethics. A correct understanding of the necessity and possibility of judicial ethics, and the promotion of judicial ethics should also become the proper intention of optimizing the judicial environment in China.

3

AHP-based Ethical Risk Evaluation Indicator System and Models

3.1

Improvement of the AHP method

In this paper, the improved analytic hierarchy process (AHP) and fuzzy comprehensive evaluation method are used to establish a comprehensive quantitative evaluation model of ethical risk. The improved AHP method is improved on the basis of the traditional AHP method, simplifying the original complex calculation process, although the nine-scale method is carefully divided, but the selectivity is not direct and effective for the survey subjects in this study: the first step: replace the “nine-scale method” with the “three-scale method”, and when comparing the two pairs, “2” represents “A is more important than B”, “0” means “B is more important than A”, and “1” represents “A is as important as B”, and the comparison matrix is constructed. Step 2: Calculate the optimal consistency matrix. Step 3: The eigenvalue vectors calculated by the optimization consistency matrix are arranged in order of size to obtain the weights of each factor combination.

The final results obtained do not require a consistency test, and the principle of scientific simplicity is met while achieving the purpose of the calculation.

The specific steps of the improved AHP method are described below.

Step 1: Establish the hierarchy system.

Step 2: Using the three scales (2, 0, 1), construct the required comparison matrix A_ij.

Step 3: Using equation (1), transform to construct the real judgment matrix B. (1) ${\begin{matrix} a_{i j} = \frac{r_{i} - r_{j}}{r_{\max} - r_{\min}} \times (b_{m} - 1) + 1, r_{i} \geq r_{j} \\ a_{i j} = 1, r_{i} = r_{j} \\ a_{i j} = {[\frac{| r_{i} - r_{j} |}{r_{\max} - r_{\min}} \times (b_{m} - 1) + 1]}^{- 1}, r_{i} < r_{j} \end{matrix}$

where $r_{i} = \sum a_{i j} r_{\max} = \max (r_{i}) r_{\min} = \min (r_{i})$ ; $b_{m} = \frac{r_{\max}}{r_{\min}}$ .

Step 4: Calculate the optimal transfer matrix C according to Eq. (2). (2) $c_{i j} = \frac{1}{n} \sum_{k = 1}^{n} (\lg \frac{a_{i k}}{a_{j k}})$

In the formula, ∀i, j; k = 1, 2, ⋯, n.

Step 5: Based on Eq. (3), the judgment matrix a^* is calculated. (3) $a^{*} = 10^{c_{i j}}$

Step 6: According to equation (4), the weight vector w_j for each indicator can be solved after calculating a^*. (4) $w_{j} = \frac{1}{\sum a_{i j}^{*}}$

3.2

Improvement of AHP-fuzzy comprehensive evaluation method

1)

Determine the factor set D = U₁, U₂, ⋯, U_n of the evaluation object, and its sub-factor set is U_i = u_i1, u_i2, ⋯, u_in.

2)

Construct the weight vectors according to the improved AHP. The weight vectors corresponding to U are $A = (a_{1}, a_{2}, \dots, a_{n})$ , i = 1, 2, ⋯, n. The weight vectors corresponding to U_i are $A_{i} = (a_{i 1}, a_{i 2}, \dots, a_{i j}, \dots, a_{i m})$ , i = 1, 2, ⋯, m.

3)

Determine the rating of the rubric $V = (v_{1}, v_{2}, \dots, v_{n})$ , the rubric set used in this study is graded in 5 levels: good fairness, good, fair, poor, very poor, i.e. n = 5.

4)

Perform a one-factor fuzzy evaluation, which can be calculated to produce a one-factor evaluation matrix. (5) $R = [\begin{matrix} r_{11} & r_{12} & \dots & r_{1 m} \\ r_{21} & r_{22} & \dots & r_{2 m} \\ ⋮ & ⋮ & ⋮ & ⋮ \\ r_{n 1} & r_{n 2} & \dots & r_{n m} \end{matrix}], 0 \leq r_{i j} \leq 1$

5)

Construct first level, second level fuzzy for transformation. (1)

First-level fuzzy comprehensive judgment. The weight vector corresponding to U_i is $A_{i} = (a_{i 1}, a_{i 2}, \dots, a_{i j}, \dots, a_{i m})$ , and the result can be obtained as B_i = A_iR_i.

(2)

Second-level fuzzy comprehensive evaluation. According to the first-level fuzzy comprehensive evaluation can be obtained (6) $B = [\begin{matrix} B_{1} \\ B_{2} \\ ⋮ \\ B_{i} \end{matrix}]$

4

Example analysis of the application of the Numerical Intelligence Forensic Ethics Risk Evaluation Model

4.1

Establishment of a system of ethical risk evaluation indicators

Taking into account the characteristics of the digitalization of forensic science and the principles of comprehensiveness, systematicity and scientificity that should be adhered to in the selection of indicators, the results of the identification of stakeholders and the analysis of their legitimacy and urgency are used to identify the ethical risk factors at each stage. In the four stages of demand identification, technology implementation, result evaluation, and decision execution, judges, forensic experts, and parties are identified as stakeholders, and abuse of power by judges, insufficient application rights by parties, technological limitations ignored by forensic experts, uncontrolled use of technology by forensic experts, disclosure of private data by forensic experts in the process of using AI, restriction of the parties’ right to know, biased examination by judges, inaccurate review by forensic experts, misjudgment by judges, and inaccuracy of parties due to machines are identified. The 10 specific risks are: inaccurate review by the expert, misjudgment by the judge, and insufficient trust index due to machine bias of the parties. The systematic framework of “Judicial Ethics Risk Indicators - 4 Stages - Key Stakeholders in Each Stage - Specific Risks” is used to establish the evaluation framework. The systematic framework of “Judicial Ethics Risk Indicators - 4 Stages - Key Stakeholders at Each Stage - Specific Risks” is used to establish the evaluation indicator system, and the specific indicators are divided as shown in Table 1.

Table 1.

Index system of ethical risk evaluation

Evaluation index system of judicial ethics(A)	Project phase	Stakeholder	Specific factors of ethical risk
	Requirement identification(B1)	Judge	Abuse of power(C1)
	Requirement identification(B1)	Party	Insufficient right of application(C2)
	Technology implementation(B2)	Judicial expert	Ignoring Technical Limitations(C3)
			Uncontrolled use of technology(C4)
			Disclosure of private data(C5)
		Party	The right to know is limited(C6)
	Outcome evaluation(B3)	Judge	Biased review(C7)
	Outcome evaluation(B3)	Judicial expert	The review is not accurate(C8)
	Execution of judgment(B4)	Judge	Misjudge(C9)
	Execution of judgment(B4)	Party	Deficient trust index(C10)

4.2

Calculation of the weights of the ethical risk evaluation indicators

4.2.1

Subjective weight calculation and consistency test

The judgment matrix was constructed based on the improved AHP method. For the ethical risks involved in digital forensics, 10 experts were invited to score using the three-scaled scale method, and the experts concerned were all from universities, engaged in the theoretical research of judicial ethics, and had been practicing for 3 to 12 years. According to the scoring results, the relative importance of each indicator was determined, and the judgment matrix was constructed. And through the Exce1 software weight calculation, we get the weights of indicators at all levels, and the results of the weight calculation of each indicator are shown in Table 2, and the subjective weights of the risk of demand identification B1, the risk of technical implementation B2, the risk of result evaluation B3, and the risk of adjudication execution B4 are 0.154, 0.437, 0.305, and 0.104, respectively.

Table 2.

Normalized weight value and ranking of ethical risk evaluation indicators

Criterion layer	Weight	Index level	Weight	Overall target weight	Sort
B1	0.154	C1	0.518	0.080	6
B1	0.154	C2	0.482	0.074	8
B2	0.437	C3	0.345	0.151	2
		C4	0.238	0.104	5
		C5	0.245	0.107	4
		C6	0.172	0.075	7
B3	0.305	C7	0.368	0.112	3
B3	0.305	C8	0.632	0.193	1
B4	0.104	C9	0.387	0.040	10
B4	0.104	C10	0.613	0.064	9

The average random consistency test was performed on this judgment matrix indicator, and the statistical results of each value are shown in Table 3, which shows that CR ≤ 0.1, and all indicators passed the consistency test.

Table 3.

Statistics of each value

	λ_max	n	CI	RI	CR
A	4.023	4	0.008	0.9	0.0089
B1	2.011	2	0.006	0	0
B2	4.009	4	0.005	0.9	0.0056
B3	2	2	0	0	0
B4	2.017	2	0.009	0	0

4.2.2

Determination of combined weights of indicators

This paper applies the entropy value method to objectively assign weights to the ethical risk evaluation indexes, with the help of the entropy theory of intuitionistic fuzzy set to calculate the objective weights of the indexes. According to the previous analysis, this paper firstly needs to transform the evaluation matrix into an entropy matrix, and then assign the corresponding expert weight to each entropy value to get the entropy matrix based on expert weight. Finally, the column vectors of the above entropy matrix are summed up to get the entropy value of each indicator, and then the entropy value weight transformation formula is applied to calculate the objective weight of each evaluation indicator.

After calculating the subjective weight $W_{i}^{*}$ and objective weight $W_{i}^{A}$ of the indicators, the comprehensive weight W_i of each indicator can be obtained by the following equation, and the calculation results are shown in Table 4, the weights of the demand identification risk B1, technology implementation risk B2, result evaluation risk B3, and referee execution risk B4 are 0.168, 0.477, 0.253, and 0.102, respectively. (7) $W_{i} = 0.5 W_{i}^{*} + 0.5 W_{i}^{Δ}, (i = 1, 2, \dots, 20)$

Table 4.

Index weight calculation results

Criterion layer	Weight	Index level	Subjective weight	Objective weight	Comprehensive weight
B1	0.168	C1	0.080	0.073	0.077
B1	0.168	C2	0.074	0.109	0.091
B2	0.477	C3	0.151	0.198	0.174
		C4	0.104	0.113	0.108
		C5	0.107	0.112	0.110
		C6	0.075	0.095	0.085
B3	0.253	C7	0.112	0.077	0.095
B3	0.253	C8	0.193	0.122	0.158
B4	0.102	C9	0.040	0.042	0.041
B4	0.102	C10	0.064	0.059	0.061

4.3

Comprehensive Ethical Risk Evaluation Model Construction

In this paper, the case of Numerical Intelligence Forensics X in City A in 2024 is selected to be analyzed and its ethical risk is assessed.

The comprehensive evaluation model of the ethical risk of Numerical Forensics is mainly composed of two parts: the weight of the indicators and the risk value of the indicators. After determining the weight of the indicator layer using the weighting model, it is also necessary to measure the risk value of each indicator layer. The risk value of the ethical risk indicator layer can be expressed in terms of the probability of occurrence of the ethical risk $(P)$ and the loss consequences of the risk $(L)$ , with the following expression: (8) $R^{″} = F (P, L)$

Where: R′′ is the risk value of the ethical risk indicator layer: F represents the functional equation of the ethical risk indicator layer; P is the probability of occurrence of the ethical risk; and L is the consequence of loss resulting from the ethical risk.

To determine the risk value R′′ of the indicator layer, it is first necessary to determine the specific values of the two parameters: the probability of ethical risk occurring $(P)$ and the consequences of the loss generated by the risk $(L)$ . Since these two parameters are qualitative factors that cannot be directly quantified by collecting existing data, but can only be quantified through the experience of experts, the data collection of these two parameters in this paper is carried out by way of expert evaluation. At the same time, considering the introduction of intuitionistic fuzzy set theory, experts need to carry out forward and reverse scoring twice to form the intuitionistic fuzzy assessment number $〈 u (x), v (x) 〉$ when evaluating.

4.3.1

Risk evaluation classification

The probability of ethical risk occurrence is proportional to the risk value of the indicator, i.e., the higher the probability of ethical risk occurrence, the higher its risk value. In this paper, the ethical risk evaluation level of Numerical Forensics is defined by equal scores in the interval of $[0, 1]$ according to the commonly used five-level rubric set of the probability of risk occurrence {low risk, low risk, medium risk, high risk, high risk}, and the specific descriptions of these five risk evaluation levels are shown in Table 5.

Table 5.

Detailed explanation of occurrence probability of ethical risk

The probability of a risk occurring	Assignment interval	Specify
Low	[0,0.2)	None occurred within the scope of the assessment, and similar projects are rare
Relatively low	[0.2,0.4)	None occurred within the scope of the assessment, and similar projects occur occasionally
Middling	[0.4,0.6)	Did not occur within the scope of assessment, but similar projects occur from time to time; It has occurred within the scope of the assessment, and similar projects occasionally occur
Relatively high	[0.6,0.8)	It happened within the scope of the evaluation, and similar projects happen from time to time
High	[0.8,1)	It has occurred within the scope of the assessment, and similar projects often occur

4.3.2

Ethical risk value calculation

After distributing the risk evaluation forms to the experts, it is necessary to summarize and organize these risk evaluation forms to calculate the risk value of the evaluation indicator. In this paper, 10 risk evaluation forms were issued and 10 were recovered, therefore m = 10. By summarizing the data of these 10 risk evaluation forms, the probability of occurrence of ethical risks of evaluation indicators P and the consequences of losses generated by the ethical risks of evaluation indicators L were calculated, and after calculating the values of the two variables for all indicator layers $(P, L)$ , the value of the risk of the indicator strata ${R^{″}}_{i}$ was obtained by calculating the value of the risk of the indicator strata according to the formula (9), and the results of the calculations are shown in Table 6. (9) $R_{i}^{″} = \sqrt{\frac{P_{i}^{2} + L_{i}^{2}}{2}}$

Table 6.

Calculation results of expert scores

Index level	Expert rating calculation results
Index level	P	L	${R^{″}}_{i}$
C1	0.484	0.386	0.438
C2	0.197	0.298	0.253
C3	0.922	0.892	0.907
C4	0.797	0.784	0.791
C5	0.722	0.938	0.837
C6	0.686	0.841	0.767
C7	0.611	0.592	0.602
C8	0.928	0.805	0.869
C9	0.602	0.308	0.478
C10	0.244	0.389	0.325

After determining the risk value of the indicator layer ${R^{″}}_{i}$ , the calculated comprehensive weight of the indicators, i.e., the absolute weight, is further calculated to obtain the absolute risk value of the ethical risk guideline layer of the Numerical Intelligence Forensics, and the cumulative sum can be obtained to obtain the ethical risk value of the whole Numerical Intelligence Forensics program, and the calculation results are shown in Table 7. According to the results of the Numerical Forensics Ethical Risk Comprehensive Evaluation Model, the overall ethical risk value of the project is 0.691.

Table 7.

Results of ethical risk assessment

Overall Objective	Criterion layer	Index level	Comprehensive weight (W_i)	Value at risk ${R^{″}}_{i}$	Weighted risk value $(W_{i} * {R^{″}}_{i})$
A (0.691)	B1(0.057)	C1	0.077	0.438	0.034
	B1(0.057)	C2	0.091	0.253	0.023
	B2(0.400)	C3	0.174	0.907	0.158
		C4	0.108	0.791	0.085
		C5	0.110	0.837	0.092
		C6	0.085	0.767	0.065
	B3(0.194)	C7	0.095	0.602	0.057
	B3(0.194)	C8	0.158	0.869	0.137
	B4(0.040)	C9	0.041	0.478	0.020
	B4(0.040)	C10	0.061	0.325	0.020

4.3.3

Comprehensive Ethical Risk Rating Evaluation

The results of the comprehensive risk level evaluation obtained against the ethical risk evaluation levels delineated in Table 5 are shown in Table 8. The ethical risk value of the project is in the range of [0.6, 0.8), and its ethical risk evaluation level is high risk. The ethical risk value of the technology implementation is in the interval of [0.4, 0.6), and its ethical risk evaluation level is medium risk. It means that the ethical risks that may be encountered by Numerical Intelligence Forensics may occur and cause certain losses, among which the ethical risk of technology implementation should be emphasized, and for this reason, it is necessary to establish an effective ethical risk prevention system to prevent and resolve ethical risks, in order to reduce the occurrence of ethical problems.

Table 8.

Comprehensive risk grade evaluation

Criterion layer	Weighted risk value	The probability of a risk occurring	Index level	Weighted risk value	The probability of a risk occurring
B1	0.057	Low	C1	0.034	Low
B1	0.057	Low	C2	0.023	Low
B2	0.400	Middling	C3	0.158	Low
			C4	0.085	Low
			C5	0.092	Low
			C6	0.065	Low
B3	0.194	Low	C7	0.057	Low
B3	0.194	Low	C8	0.137	Low
B4	0.040	Low	C9	0.020	Low
B4	0.040	Low	C10	0.020	Low

5

Conclusions and strategies

5.1

Conclusion

This paper constructs a numerical intelligence forensic ethical risk evaluation model based on the improved AHP-fuzzy comprehensive evaluation method, and analyzes the application of X case as an example.

The weights of demand identification risk B1, technology implementation risk B2, result assessment risk B3, and adjudication execution risk B4 are 0.168, 0.477, 0.253, and 0.102, respectively. The overall ethical risk value of X case is 0.691, and the ethical risk value is in the interval of [0.6, 0.8), and its ethical risk evaluation level is higher risk. The ethical risk value of technology implementation is 0.4, which is within the interval of [0.4, 0.6), and its ethical risk evaluation level is medium risk. It means that the ethical risks that may be encountered by Numerical Intelligence Forensics may occur and cause certain losses, among which the ethical risk of technology implementation should be emphasized.

5.2

Strategies

Aiming at the ethical risks of digital forensics, this paper proposes the legal and ethical supervision methods of digital forensics from the three levels of technical governance, institutional regulation and collaborative supervision. 1)

Technical governance level

The law should stipulate that all algorithmic models applied to forensic identification undergo public review and evaluation to ensure that the models are impartial and unbiased. Regularly conduct technical review and validation, especially validation and ethical assessment of critical algorithmic models. Establish a technical committee dedicated to ethical review and decision-making on digital forensic technology.

2)

Institutional regulation level

Formulate specialized laws and regulations related to digital forensics to clarify the ethical requirements in the application of the technology. Ensure that the entire process of data collection, storage, analysis and use involved in forensic identification is protected and regulated by law. Incorporate ethical risk assessment into the standardized process of forensic appraisal and ensure that every digital forensic activity must be preceded by an ethical risk assessment. Establish a clear mechanism for responsibility tracing, so that both the technology developer, judicial personnel, and all parties involved in the appraisal should be responsible for the ethical risks of their behavior.

3)

Collaborative Supervision Level

Establish a third-party independent oversight mechanism to provide external oversight of digital forensics through civic institutions, academic organizations, and professional groups to prevent technological abuse or dereliction of duty. Third-party independent review helps to enhance trust in the technology and provide professional advice for system improvement.

The improvement of ethical risk prevention and control mechanisms not only safeguards the impartiality and legitimacy of forensic identification, but also promotes public trust and acceptance of the judicial system.

Sprache:: Englisch

Zeitrahmen der Veröffentlichung:: 1 Hefte pro Jahr
Fachgebiete der Zeitschrift:: Biologie, Biologie, andere, Mathematik, Angewandte Mathematik, Mathematik, Allgemeines, Physik, Physik, andere

Zeitschrift RSS Feed

Exploring the Ethical and Legal Boundaries of Artificial Intelligence in Forensic and Document Examination

Jing Ye

Haoying Du

Online veröffentlicht: 29. Sept. 2025

Eingereicht: 11. Jan. 2025

Akzeptiert: 11. Mai 2025

DOI: https://doi.org/10.2478/amns-2025-1126

SchlüsselwörterDigital intelligence appraisal, Judicial ethics, AHP method, Intuitionistic fuzzy set theory, Risk assessment

© 2025 Jing Ye and Haoying Du, published by Sciendo.

This work is licensed under the Creative Commons Attribution 4.0 International License.

Schlüsselwörter
Digital intelligence appraisal, Judicial ethics, AHP method, Intuitionistic fuzzy set theory, Risk assessment