Skip to main content
Log in

Subjective logic operators in trust assessment: an empirical study

  • Published:
Information Systems Frontiers Aims and scope Submit manuscript

Abstract

Computational trust mechanisms aim to produce trust ratings from both direct and indirect information about agents’ behaviour. Subjective Logic (SL) has been widely adopted as the core of such systems via its fusion and discount operators. In recent research we revisited the semantics of these operators to explore an alternative, geometric interpretation. In this paper we present principled desiderata for discounting and fusion operators in SL. Building upon this we present operators that satisfy these desirable properties, including a family of discount operators. We then show, through a rigorous empirical study, that specific, geometrically interpreted, operators significantly outperform standard SL operators in estimating ground truth. These novel operators offer real advantages for computational models of trust and reputation, in which they may be employed without modifying other aspects of an existing system.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4

Similar content being viewed by others

Notes

  1. An example of this is GPS data, which is known to be inaccurate if you are using civilian equipment (Bisdikian et al. 2012).

  2. Hereafter each opinion will have a fixed relative atomicity of \(\frac {1}{2}\). This assumption will be relaxed in future work.

  3. (Jøsang et al. 2005) shows how the deduction operator affects the space of possible derived opinions.

  4. This representation is similar to the one used in Jøsang (2001) for representing opinions in SL, but points B and D are swapped.

  5. With reference to Fig. 1, \(\alpha _{C} \triangleq {\angle }_{CBD}\), \(\beta _{T} \triangleq \angle _{TDB}\), \(\gamma _{T} \triangleq \angle _{TDU}\), \(\delta _{T} \triangleq \angle _{TUD}\), \(\epsilon _{T} \triangleq \angle _{DTU}\).

  6. The term “connected” here can have different names in different contexts, like “friend” in Facebook, or “follower” in Twitter.

  7. Although this may seem counter-intuitive, it partially captures real-world social media. For instance, Twitter messages are public, therefore we do not know who will read our messages. The same applies with slight modifications to Google+, and, of course, to blogging activities in general.

  8. Figs. 2a, b, 3a–h, are obtained using the Matlab function histfit.

  9. In Sections 7.1 and 7.2 we show that the distances are not normally distributed and thus from a statistical point of view medians rather than means should be considered. Here, however, we are more interested in the qualitative dynamics of values obtained by varying the parameters of the experiment, and thus we rely on graphical representation of mean and standard deviation.

  10. 10 The code used for this experiment is available at https://sourceforge.net/projects/slef/.

References

  • Bisdikian, C., Sensoy, M., Norman, T.J., Srivastaval, M.B. (2012). Trust and obfuscation principles for quality of information in emerging pervasive environments. In Proceedings of international workshop on information quality and quality of service for pervasive computing (IQ2S) (pp. 44–49). IEEE. doi: 10.1109/PerComW.2012.6197532.

  • Burnett, C., & Oren, N. (2012). Sub-delegation and trust. In Proceedings of the 11th international conference on Autonomous agents and multiagent systems (pp. 1359–1360). IFAAMAS. http://dl.acm.org/citation.cfm?id=2343776.

  • Castelfranchi, C., & Falcone, R. (2010). Trust theory: A socio-cognitive and computational model. Wiley Series in Agent Technology.

  • Cerutti, F., Toniolo, A., Oren, N., Norman, T. (2013a). An empirical evaluation of geometric subjective logic operators. In C.I. Chesñevar, E. Onaindia, S. Ossowski, G. Vouros (Eds.), Proceedings of the 2nd conference on agreement technologies. Lecture Notes in Computer Science (Vol. 8068, pp. 90–104). Berlin Heidelberg New York: Springer.

  • Cerutti, F., Toniolo, A., Oren, N., Norman, T.J. (2013b). Context-dependent trust decisions with subjective logic. Tech. Rep. ABDNCS 20130501, Department of Computing Science, University of Aberdeen. http://homepages.abdn.ac.uk/f.cerutti/pages/papers/techreps/ABDN-CS-2013-05-01.pdf.

  • Dempster, A.P. (1968). A Generalization of Bayesian Inference. Journal of the Royal Statistical Society Series B (Methodological), 30 (2), 205–247.

    Google Scholar 

  • Guha, R., Kumar, R., Raghavan, P., Tomkins, A. (2004). Propagation of trust and distrust. In Proceedings of the 13th conference on world wide web (pp. 403–412). New York: ACM Press. http://dl.acm.org/citation.cfm?id=988727.

  • Haghpanah, Y., & des Jardins, M. (2012). Prep: A probabilistic reputation model for biased societies. In Proceedings of the 11th international conference on autonomous agents and multiagent systems (pp. 315–322). IFAAMAS. http://dl.acm.org/citation.cfm?id=2343776.

  • Ismail, R., & Jøsang, A. (2002). The beta reputation system. In Proceedings of the 15th bled electronic commerce conference (pp. 324–337). https://domino.fov.uni-mb.si/ECOMFrames.nsf/pages/bled2002.

  • Jøsang, A. (2001). A logic for uncertain probabilities. International Journal of Uncertainty, Fuzziness and Knowledge-Based Systems, 9 (3), 279–311.

    Article  Google Scholar 

  • Jøsang, A. (2008). Cumulative and averaging unfusion of beliefs. In Proceedings of the international conference on information processing and management of uncertainty.

  • Jøsang, A., & McAnally, D. (2004). Multiplication and comultiplication of beliefs. International Journal of Approximate Reasoning, 38 (1), 19–51.

    Article  Google Scholar 

  • Jøsang, A., Pope, S., Daniel, M. (2005). Conditional deduction under uncertainty. In L. Godo (Ed.), Proceedings of the symbolic and quantitative approaches to reasoning with uncertainty. 8th European conference, ECSQARU 2005, July 6–8, 2005, Lecture Notes in Computer Science. Barcelona, Spain: Springer. ISBN 3-540-27326-3.

  • Jøsang, A., Marsh, S., Pope, S. (2006). Exploring different types of trust propagation In K. Stølen, W.H. Winsborough, F. Martinelli, F. Massacci (Eds.), Trust management (pp. 179–192). Berlin Heidelberg New York: Springer.

  • Jøsang, A., Ismail, R., Boyd, C. (2007). A survey of trust and reputation systems for online service provision. Decision Support Systems, 43 (2), 618–644.

    Article  Google Scholar 

  • Jøsang, A., Azderska, T., Marsh, S. (2012). Trust transitivity and conditional belief reasoning. In T. Dimitrakos, R. Moona, D. Patel, D.H. McKnight (Eds.), Proceedings of 6th IFIPTM international conference on trust management (pp. 68–83). Berlin Heidelberg New York: Springer.

  • McAnally, D., & Jøsang, A. (2004). Addition and subtraction of beliefs. In Proceedings of the conference on information processing and management of uncertainty in knowledge-based systems.

  • Sabater, J., & Sierra, C. (2005). Review on computational trust and reputation models. Artificial Intelligence Review, 24, 33–60.

    Article  Google Scholar 

  • Sensoy, M., Fokoue, A., Pan, J.Z., Norman, T.J., Tang, Y., Oren, N., Sycara, K. (2013). Reasoning about uncertain information and conflict resolution through trust revision. In Proceedings of the 2013 international conference on Autonomous agents and multi-agent systems.

  • Shafer, G. (1976). A mathematical theory of evidence. Princeton: Princeton University Press.

  • Urbano, J., Rocha, A., Oliveira, E. (2013). A socio-cognitive perspective of trust In S. Ossowski (Ed.) Agreement technologies, law, governance and technology series (Vol. 8, pp. 419–429). Berlin Heidelberg New York: Springer.

Download references

Acknowledgments

The authors thank the anonymous reviewers for their helpful comments.

Research was sponsored by US Army Research laboratory and the UK Ministry of Defence and was accomplished under Agreement Number W911NF-06-3-0001. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the US Army Research Laboratory, the U.S. Government, the UKMinistry of Defense, or the UK Government. The US and UK Governments are authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation hereon.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Federico Cerutti.

Appendices

Appendix A: The geometry of subjective logic

A SL opinion \(O \triangleq {{\langle {b_{O}}, {d_{O}}, {u_{O}} \rangle }}\) is a point in the \({\mathbb {R}}^{3}\) space, identified by the coordinate b O for the first axis, d O for the second axis, and u O for the third axis. However, due to the requirement that b O +d O +u O =1, an opinion is a point inside (or at least on the edges of) the triangle \({\mathop {BDU}\limits ^{\triangle }}\) shown in Fig. 5, where B=〈1,0,0〉,D=〈0,1,0〉,U=〈0,0,1〉.

Fig. 5
figure 5

The subjective logic plane region

Definition 14

The Subjective Logic plane region \({\overset {\triangle }{BDU}}\) is the triangle whose vertices are the points \(B \triangleq {\langle 1,0,0 \rangle }\), \(D\triangleq {\langle 0,1,0 \rangle }\), and \(U\triangleq {\langle 0,0,1 \rangle }\) on a \({\mathbb {R}}^{3}\) space where the axes are respectively the one of belief, disbelief, and uncertainty predicted by SL.

Since an opinion is a point inside triangle \({\mathop {BDU}\limits ^{\triangle }}\), it can be mapped to a point in Fig. 6. This representation is similar to the one used in Jøsang (2001) for representing opinions in SL, but here the belief and disbelief axes are swapped.

Fig. 6
figure 6

An opinion \(O \triangleq {{\langle {b_{O}}, {d_{O}}, {u_{O}} \rangle }}\) in SL after the \(1:\frac {\sqrt {3}}{\sqrt {2}}\) scale. The belief axis is the line from B 0 (its origin) toward the B vertex, the disbelief axis is the line from D 0 toward the D vertex, and the uncertainty axis is the line from U 0 toward the U vertex

In order to keep the discussion consistent with Jøsang’s work (Jøsang 2001), in what follows we will scale triangle \({\mathop {BDU}\limits ^{\triangle }}\) by a factor \(1:\frac {\sqrt {3}}{\sqrt {2}}\) thus obtaining that \(|\overrightarrow {B_{0} B}| = |\overrightarrow {D_{0} D}| = |\overrightarrow {U_{0} U}| = 1\).

These geometric relations lie at the heart of the Cartesian transformation operator, which is the subject of the next subsection.

A.1 The cartesian representation of opinions

As shown in Appendix A, an opinion in SL can be represented as a point in a planar figure (Fig. 6) laying on a Cartesian plane. In this section we will introduce the Cartesian transformation operator, which returns the Cartesian coordinate of an opinion.

First of all, let us define the axes of the Cartesian system we will adopt.

Definition 15

Given the SL plane region \(\mathop {BDU}\limits ^{\triangle }\) , the associated Cartesian system is composed by two axes, named respectively x,y, where the unit vector of the x axis \({\overrightarrow {e_{x}}} = \frac {1}{|\overrightarrow {BD}|} \overrightarrow {BD}\), the unit vector of the y axis the y axis \({\overrightarrow {e_{y}}} = {\overrightarrow {e_{{u_{}}}}}\), and B is the origin.

Figure 7 depicts this Cartesian system.

Fig. 7
figure 7

An opinion and its representation in the Cartesian system

The correspondence between the three values of an opinion and the corresponding coordinate in the Cartesian system we defined is shown in the following proposition.

Proposition 4

(Cerutti et al. 2013b, Proposition 1) Given a SL plane region \(\mathop {BDU}\limits ^{\triangle }\) and its associated Cartesian system 〈x,y〉, an opinion \(O \triangleq {{\langle {b_{O}}, {d_{O}}, {u_{O}} \rangle }}\) is identified by the coordinate 〈x O ,y O 〉 s.t.:

  • \({x_{O} \triangleq \frac {{d_{O}} + {u_{O}}~\cos (\frac {\pi }{3})}{\sin (\frac {\pi }{3})}};\)

  • \(y_{O} \triangleq {u_{O}}.\)

Proof

Proving that \(y_{O} \triangleq {u_{O}}\) is trivial.

Let us focus on the first part of the proposition. Consider Fig. 7. Given O, we note that the for the point P, \(\frac {1}{|\overrightarrow {P O}|} \overrightarrow {P O} = {\overrightarrow {e_{b}}}\) (i.e. \(\overrightarrow {PO}\) is parallel to the disbelief axis) and \(\frac {1}{|\overrightarrow {BP}|} \overrightarrow {BP} = \frac {1}{|\overrightarrow {BU}|} \overrightarrow {BU}\) (i.e. P is on the line \(\overrightarrow {BP}\)), and therefore \({\angle _{BPO}} = \frac {\pi }{2}\). Then we must determine Q and R s.t. \(\overrightarrow {Q R} = \overrightarrow {P O}\) and y R =0. By construction \(|\overrightarrow {P O}| = |\overrightarrow {Q R}| = {d_{O}}\), y R =0. By construction \(|\overrightarrow {P O}| = |\overrightarrow {Q R}| = {d_{O}}\), \({\angle _{Q R B}} = \frac {\pi }{6}\), \({\angle _{ORD}} = \frac {\pi }{3}\), and \(x_{O} \triangleq |\overrightarrow {BS}| = |\overrightarrow {BR}| + |\overrightarrow {RS}|\), where \(|\overrightarrow {BR}| = \frac {{d_{O}}}{\sin (\frac {\pi }{3})}\), and \(|\overrightarrow {RS}| = \frac {{u_{O}}}{\sin (\frac {\pi }{3})} \cos (\frac {\pi }{3})\). □

There are some notable elements of Fig. 7 that we will repeatedly use below, and we therefore define them as follows:

  • the angle α O determined by the x axis and the vector \(\overrightarrow {BO}\);

  • the three angles (γ O ,δ O , and 𝜖 O ) of the triangle \({\overset {\triangle }{ODU}}\), namely the triangle determined by linking the point O with the vertex D and U through straight lines.

Definition 16

Given the SL plane region \(\mathop {BDU}\limits ^{\triangle }\), given O=〈b O ,d O ,u O 〉 whose coordinates are 〈x O ,y O 〉 where \({x_{O} \triangleq \frac {{d_{O}} + {u_{O}}~\cos (\frac {\pi }{3})}{\sin (\frac {\pi }{3})}}\) and \(y_{O} \triangleq {u_{O}}\), let us define and (via trivial trigonometric relations) compute the following.

  • \({\alpha _{O} \triangleq {\angle _{OBD}}} = \)

    \(\left \{ \begin {array}{l l} 0 & \text {if } {b_{O}} = 1\\ {\arctan \left (\frac {{u_{O}}~\sin (\frac {\pi }{3})}{{d_{O}} + {u_{O}}~\cos (\frac {\pi }{3})}\right )} & \text {otherwise} \end {array} \right .\);

  • \({\beta _{O} \triangleq {\angle _{ODB}} } = \)

    \(\left \{ \begin {array}{l l} {\frac {\pi }{3}} & \text {if } {d_{O}} = 1\\ {\arctan \left (\frac {{u_{O}}~\sin (\frac {\pi }{3})}{1-({d_{O}}+{u_{O}}~\cos (\frac {\pi }{3}))}\right )} & \text {otherwise} \end {array}\right .\);

  • \({\gamma _{O} \triangleq {\angle _{ODU}} = \frac {\pi }{3}} - \beta _{O}\);

  • \({\delta _{O} \triangleq {\angle _{OUD}} = }\)

    \(\left \{ \begin {array}{l l} 0 & \text {if } {u_{O}} = 1\\ {\arcsin \left (\frac {{b_{O}}}{|\overrightarrow {OU}|}\right )} & \text {otherwise} \end {array}\right .\);

  • \(\epsilon _{O} \triangleq {\angle _{DOU}} = \pi - \gamma _{O} - \delta _{O}\);

where \({|\overrightarrow {OU}| = \sqrt {\frac {1}{3} (1 + {d_{O}} - {u_{O}})^{2} + {b_{O}}^{2}}}\).

The angle α O is called the direction of O.

Equivalently, we can write \(\overrightarrow {BO}\) or \({\langle B, \alpha _{O}, |\overrightarrow {BO}| \rangle }\).

Finally, as an element of SL is bounded to have its three components between 0 and 1, we are also interested in determining the point M O such that the vector \(\overrightarrow {BM_{O}}\) has the maximum magnitude given (a) the direction α O of an opinion O, and (b) M O is a SL opinion. In other words, determining the magnitude of \(\overrightarrow {BM_{O}}\) will allow us to re-define the vector \(\overrightarrow {BO}\) as a fraction of \(\overrightarrow {BM_{O}}\).

Definition 17

Given the SL plane region \({\mathop {BDU}\limits ^{\triangle }}\), and \(O \triangleq {{\langle {b_{O}}, {d_{O}}, {u_{O}} \rangle }}\) whose coordinates are 〈x O ,y O 〉 where \({x_{O} \triangleq \frac {{d_{O}} + {u_{O}}~\cos (\frac {\pi }{3})}{\sin (\frac {\pi }{3})}}\) and \(y_{O} \triangleq {u_{O}}\), and \({\alpha _{O} \triangleq {\angle _{OBD}} = \arctan \left (\frac {{u_{O}}~\sin (\frac {\pi }{3})}{{d_{O}} + {u_{O}}~\cos (\frac {\pi }{3})}\right )}\), let us define \(M_{O} \triangleq {\langle x_{M_{O}}, y_{M_{O}} \rangle }\) as the intersection of the straight line passing for O and B, and the straight line passing for U and D, and thus define the following.

  • \(x_{M_{O}} \triangleq \frac {2 - y_{O} + \tan (\alpha _{O})~ x_{O}}{\tan (\alpha _{O}) + \sqrt {3}}\);

  • \({y_{M_{O}} \triangleq -\sqrt {3}~ x_{M_{O}} + 2}\).

Appendix B: Proof

Proof

  • i. Let us prove that 0≤b W ≤1, 0≤d W ≤1, 0≤u W ≤1.

    To prove that \({b_{{W_{}}}} \geq 0\), \({d_{{W_{}}}} \geq 0\), and \({u_{{W_{}}}} \geq 0\) is trivial since Cand Tare opinions.

    \({b_{{W_{}}}} = {b_{{C_{}}}} \cdot {b_{{T_{}}}} \leq 1\) is immediate since Cand Tare opinions.

    \({d_{{W_{}}}} = {b_{}} \cdot {d_{{T_{}}}} + {d_{{C_{}}}} \leq 1\) can be rewritten as \({d_{{T_{}}}} \leq 1 + \frac {{u_{{C_{}}}}}{{b_{{C_{}}}}}\) if \({b_{{C_{}}}} \neq 0\), or \({d_{{C_{}}}} \leq 1\) otherwise. Both in-equations are verified since Cand Tare opinions.

    \({u_{{W_{}}}} = {b_{}} \cdot {u_{{T_{}}}} + {u_{{C_{}}}} \leq 1\) can be rewritten as \({d_{{T_{}}}} \leq 1 + \frac {{d_{{C_{}}}}}{{b_{{C_{}}}}}\) if \({b_{{C_{}}}} \neq 0\), or \({u_{{C_{}}}} \leq 1\) otherwise. Both in-equations are verified since Cand Tare opinions.

    Finally, \({b_{{W_{}}}} + {d_{{W_{}}}} + {u_{{W_{}}}} = {b_{{C_{}}}} ({b_{{T_{}}}} + {d_{{T_{}}}} + {u_{{T_{}}}}) + {d_{{C_{}}}} + {u_{{C_{}}}} = 1\)

  • ii.Given C=〈1,0,0〉, W=T n Cis such that:

    • \({b_{{W_{}}}} = {b_{{C_{}}}} \cdot {b_{{T_{}}}} = {b_{{T_{}}}}\);

    • \({d_{{W_{}}}} = {b_{{C_{}}}} \cdot {d_{{T_{}}}} + {d_{{C_{}}}} = {d_{{T_{}}}}\);

    • \({u_{{W_{}}}} = {b_{{C_{}}}} \cdot {u_{{T_{}}}} + {u_{{C_{}}}} = {u_{{T_{}}}}\).

  • iii.Given C=〈0,0,1〉, W=T n Cis such that:

    • \({b_{{W_{}}}} = {b_{{C_{}}}} \cdot {b_{{T_{}}}} = 0 = {b_{{C_{}}}}\);

    • \({d_{{W_{}}}} = {b_{{C_{}}}} \cdot {d_{{T_{}}}} + {d_{{C_{}}}} = {d_{{C_{}}}}\);

    • \({u_{{W_{}}}} = {b_{{C_{}}}} \cdot {u_{{T_{}}}} + {u_{{C_{}}}} = {u_{{C_{}}}}\).

  • iv.By contradiction, \({b_{{W_{}}}} = {b_{{C_{}}}} \cdot {b_{{T_{}}}} > {b_{{T_{}}}}\) leads to \({b_{{C_{}}}} > 1\), which is impossible.

Proof

By Definition 8, \({\mathbb {O}_{{T_{}}{}}} = {\{X \in {\mathbb {O}}{} | {b_{X}} \leq {b_{{T_{}}{}}}\}}\). From Proposition 4,

$$\begin{array}{ll} {b_{X}} & \leq {b_{{T_{}}}}\\ y_{X} & \geq \sqrt{3}x_{X} + 2(1 - {b_{{T_{}}}})\\ {u_{X}} & \geq 1 - {d_{X}} - {b_{{T_{}}}} \end{array} $$

Therefore:

  • if d X =0, u X ≥1−b T (limit case \({\langle {b_{{T_{}}}}, 0, 1 - {b_{{T_{}}}} \rangle }\);

  • if u X =0, d X ≥1−b T (limit case \({\langle {b_{{T_{}}}}, 1-{b_{{T_{}}}}, 0 \rangle } = Q\)).

Proof

Proving the thesis in the limit case is trivial. In the following we will assume, without loss of generality, that \(\alpha _{C^{\prime }} \neq \frac {\pi }{2}\), \(\alpha _{C^{\prime }} \neq -\frac {\pi }{3}\), \(\alpha _{C^{\prime }} \neq \frac {2}{3} \pi \).

  • (i.)W=〈b W ,d W ,u W 〉 must respect

$$ {u_{W}} + {d_{W}} \leq 1 $$
(1)

From Definition 9 it is clear that Eq. 1 can be rewritten as follows.

$$ \begin{array}{ll} {u_{T}}{}+ {d_{T}}{}+{}\frac{r_{C}}{2} \frac{2 \sqrt{\tan^{2}(\alpha_{C^{\prime}}) + 1}}{|\tan(\alpha_{C^{\prime}}) + \sqrt{3}|} ~{b_{T}}{} ~(\sin(\alpha_{C^{\prime}}) {}+{}\sqrt{3}\cos(\alpha_{C^{\prime}})) \leq 1 \end{array} $$
(2)

In turn, using the relation \(\tan (\alpha _{C^{\prime }}) = \frac {\sin (\alpha _{C^{\prime }})}{\cos (\alpha _{C^{\prime }})}\), this can be rewritten as

$${u_{T}} + {d_{T}} + r_{C} {b_{T}} \leq 1 $$

which entails the requirement that \(r_{C} \leq \frac {1 - {u_{T}} - {d_{T}}}{{b_{T}}} = \frac {{b_{T}}}{{b_{T}}} = 1\). However, from Definition 9, we know that r C ≤1, fulfilling this requirement.

  • (ii.) C=〈1,0,0〉 implies that \(|\overrightarrow {BC}| = 0\) and thus r C =0. Therefore, from Definition 9, \({u_{W}} = {u_{T}} + \sin (\alpha _{C^{\prime }}) r_{C} |\overrightarrow {TM_{C^{\prime }}}| = {u_{T}}\) and this results also implies that d W =d T . Since Point 1 shows that W is an opinion in SL, we conclude that W=T.

  • (iii.) C=〈0,0,1〉 implies r C =1, \(\frac {\frac {\pi }{3} \epsilon _{T}}{\frac {\pi }{3}} - \beta _{T} \leq \alpha _{C^{\prime }} \leq \epsilon _{T} - \beta _{T}\), and thus \(\alpha _{C^{\prime }} = \epsilon _{T} - \beta _{T} = \epsilon _{T} - \beta _{T} = \frac {2}{3} \pi - \delta _{t}\). Therefore, we obtain that \({u_{W}} = {u_{T}} + \frac {{b_{T}}}{2} (1 + \frac {\sqrt {3}}{\tan (\delta _{T})})\). From Definition 16 and the trigonometric property that \(\tan (\arcsin (v)) = \frac {v}{\sqrt {1 - v^{2}}}\) we obtain that \({u_{W}} = {u_{T}} + \frac {{b_{T}}}{2} + \frac {\sqrt {3}}{2} \sqrt {|\overrightarrow {TU}|^{2} - {b_{T}}^{2}}\). From Definition 16 we can write:

    $$ \begin{array}{ll} {u_{W}} & = {u_{T}} + \frac{{b_{T}}}{2} + \frac{\sqrt{3}}{2} \frac{1 + {d_{T}} - {u_{T}}}{\sqrt{3}}\\ & = \frac{1}{2} (1 + {b_{T}} + {d_{T}} + {u_{T}} ) = 1 \end{array} $$
    (3)

    Similarly, \({d_{W}} = {d_{T}} + \frac {{u_{T}}}{2} - \frac {1}{2} + \frac {\sqrt {3}}{2} \frac {1}{\sin (\delta _{T})} {b_{T}} = {d_{T}} + \frac {{u_{T}}}{2} - \frac {1}{2} + \frac {\sqrt {3}}{2} |\overrightarrow {TU}|\). From Definition 16 we have

    $$ \begin{array}{ll} {d_{W}} & = {d_{T}} + \frac{{u_{T}}-1}{2} + \frac{3}{4}{b_{T}} - \frac{\sqrt{3}}{4} {b_{T}} \frac{1 + {d_{T}} - {u_{T}}}{\sqrt{3} {b_{T}}} \\ & = \frac{1}{4} (4{d_{T}} + 2 {u_{T}} - 2 + 3 {b_{T}} - 1 + {u_{T}} - {d_{T}}) = 0 \end{array} $$
    (4)

    From Eqs. 3 and 4, together with Point 1, it follows that W=〈0,0,1〉=C.

  • Suppose instead b W >b T .

$$\begin{array}{@{}rcl@{}}1 - {d_{W}} - {u_{W}} > 1 - {d_{T}} - {u_{T}} \end{array} $$
$$\begin{array}{@{}rcl@{}} {d_{W}} + {u_{W}} < {d_{T}} + {u_{T}} \end{array} $$
$$\begin{array}{@{}rcl@{}} {d_{T}} + \sin(\alpha_{C^{\prime}} + \frac{\pi}{3}) \frac{r_{C}}{\sin(\alpha_{C^{\prime}} + \frac{\pi}{3})} + {u_{T}} < {d_{T}} + {u_{T}} \end{array} $$
$$\begin{array}{@{}rcl@{}} r_{C} < 0 \end{array} $$

but 0≤r C ≤1. Quod est absurdum.

Proof

  • (i.) To prove that 〈b 1(W 1,…,W n ),d 1(W 1,…,W n ),u 1(W 1,…,W n )〉 is an opinion, we have to show that u 1(W 1,…,W n ) + d 1(W 1,…,W n ) Subscript>≤1 holds.

    $$\begin{array}{rl} {u_{{{\Gamma}_{1}}(W_{1}, \ldots, W_{n})}} + {d_{{{\Gamma}_{1}}(W_{1}, \ldots, W_{n})}} = & {\frac{1}{{\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}({u_{W_{i}}} + {d_{W_{i}}})\right)}\\ = & {\frac{1}{{\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}(1 - {b_{W_{i}}})\right)}\\ = & 1 - {\frac{1}{{\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}~{b_{W_{i}}} \right)} \end{array}$$
  • (ii.) From Proposition 4,

    $$\left\{\begin{array}{l} {x_{{{\Gamma}_{1}}(W_{1}, \ldots, W_{n})} = \frac{{d_{{{\Gamma}_{1}}(W_{1}, \ldots, W_{n})}}}{\sin(\frac{\pi}{3})} + \frac{1}{2~\sin(\frac{\pi}{3})~{\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}~ {u_{W_{i}}}\right)}\\ {x_{{{\Gamma}_{1}}(W_{1}, \ldots, W_{n})} = \frac{1}{\sin(\frac{\pi}{3}) {\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}~ \left({d_{W_{i}}} + \frac{{u_{W_{i}}}}{2}\right) \right)} \end{array}\right.$$

Thus we obtain:

$$\begin{array}{rl} {d_{{{\Gamma}_{1}}(W_{1}, \ldots, W_{n})}} = & {\sin(\frac{\pi}{3})\left(\frac{1}{\sin(\frac{\pi}{3}) {\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}~ \left({d_{W_{i}}} + \frac{{u_{W_{i}}}}{2}\right) \right) + \right. }\\ & ~~~~ {\left.- \frac{1}{2~\sin(\frac{\pi}{3})~{\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}~ {u_{W_{i}}}\right) \right)}\\ = & {\frac{1}{{\sum}_{i=1}^{n} K_{i}} \left( \left(\sum\limits_{i=1}^{n} K_{i}~ \left({d_{W_{i}}} + \frac{{u_{W_{i}}}}{2}\right)\right) - \left( \sum\limits_{i=1}^{n} \frac{{u_{W_{i}}}}{2} \right) \right)}\\ = & {\frac{1}{{\sum}_{i=1}^{n} K_{i}} \left( \sum\limits_{i=1}^{n} K_{i}~ {d_{W_{i}}}\right)} \end{array}$$

Since \({\frac {1}{{\sum }_{i=1}^{n} K_{i}} \left (\sum \limits _{i=1}^{n} K_{i}~{b_{W_{i}}} \right )} \geq 0\), then u 1(W 1,…,W n ) + dΓ1(W 1,…,W n )1 holds.

  • (iii.) Immediate from Proposition 4.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Cerutti, F., Kaplan, L.M., Norman, T.J. et al. Subjective logic operators in trust assessment: an empirical study. Inf Syst Front 17, 743–762 (2015). https://doi.org/10.1007/s10796-014-9522-5

Download citation

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s10796-014-9522-5

Keywords

Navigation