Abstract
Trust models are mechanisms that allow agents to build trust without relying on a trusted central authority. Our goal was to develop a trust model that would operate with values that humans easily understand and manipulate: qualitative and ordinal values. The result is a trust model that computes trust from experiences created in interactions and from opinions obtained from third-party agents. The trust model, termed qualitative trust model (QTM), uses qualitative and ordinal values for assessing experiences, expressing opinions and estimating trust. We treat such values appropriately; we never convert them to numbers, but merely use their relative order. To aggregate a collection of such values, we propose an aggregation method that is based on comparing distributions and show some of its properties; the method can be used in other domains and can be seen as an alternative to median and similar methods. To cope with lying agents, QTM estimates trustworthiness in opinion providers with a modified version of the weighted majority algorithm, and additionally combines trustworthiness with social links between agents; such links are obtained implicitly by observing how agents provide opinions about each other. Finally, we compare QTM against a set of well-known trust models and demonstrate that it consistently performs well and on par with other quantitative models, and in many cases even outperforms them, particularly when the number of direct experiences is low.
Similar content being viewed by others
Notes
When talking about random selections in scenarios we are referring to a uniform random selection from interval \([0,1]\).
References
Abdul-Rahman, A., & Hailes, S. (2000). Supporting trust in virtual communities. In Proceedings of the 33rd annual Hawaii international conference on system sciences.
Carbo, J., Molina, J., & Davila, J. (2003). Trust management through fuzzy reputation. International Journal of Cooperative Information Systems, 12, 135–155.
Douceur, J. R. (2002). The Sybil attack. In P. Druschel, F. Kaashoek, & A. Rowstron (Eds.), Peer-to-peer systems. Berlin: Springer.
Dubois, D., & Prade, H. (1986). Weighted minimum and maximum operations in fuzzy set theory. Information Sciences, 39, 205–210.
Gambetta, D. (1990). Can we trust trust? In D. Gambetta (Ed.), Trust: Making and breaking cooperative relations. Oxford: Basil Blackwell.
Godo, L., & Torra, V. (2000). On aggregation operators for ordinal qualitative information. IEEE Transactions on Fuzzy Systems, 8, 143–154.
Griffiths, N., Chao, K. M., & Younas, M. (2006). Fuzzy trust for peer-to-peer systems. In 26th IEEE international conference on distributed computing systems workshops, 2006 (ICDCS workshops 2006).
Herrera, F., Viedma, E. H., & Verdegay, J. L. (1997). A rational consensus model in group decision making using linguistic assessments. Fuzzy Sets and Systems, 88, 31–49.
Hoffman, K., Zage, D., & Nita-Rotaru, C. (2009). A survey of attack and defense techniques for reputation systems. ACM Computing Surveys, 42, 1–31.
Jamieson, S. (2004). Likert scales: How to (ab)use them. Medical Education, 38, 1217–1218.
Jelenc, D., Hermoso, R., Ossowski, S., & Trček, D. (2012). Alpha test-bed: A new approach for evaluating trust models. In Proceedings of the third international workshop on infrastructures and tools for multiagent systems.
Jelenc, D., Hermoso, R., Sabater-Mir, J., & Trček, D. (2013). Decision making matters: A better way to evaluate trust models. Knowledge-Based Systems, 52, 147–164. doi:10.1016/j.knosys.2013.07.016.
Jennings, N. R. (2001). An agent-based approach for building complex software systems. Communications of the ACM, 44, 35–41.
Jøsang, A., & Haller, J. (2007). Dirichlet reputation systems. In The second international conference on availability, reliability and security (ARES).
Jøsang, A., & Ismail, R. (2002). The beta reputation system. In Proceedings of the 15th Bled electronic commerce conference.
Jøsang, A., Ismail, R., & Boyd, C. (2007). A survey of trust and reputation systems for online service provision. Decision Support Systems, 43, 618–644.
Kahler, E., Rogausch, A., Brunner, E., & Himmel, W. (2008). A parametric analysis of ordinal quality-of-life data can lead to erroneous results. Journal of Clinical Epidemiology, 61, 475–480.
Kahneman, D., Slovic, P., & Tversky, A. (1982). Judgment under uncertainty: Heuristics and biases. New York: Cambridge University Press.
Kamvar, S. D., Schlosser, M. T., & Garcia-Molina, H. (2003). The eigentrust algorithm for reputation management in p2p networks. In Proceedings of the 12th international conference on World Wide Web.
Knapp, T. R. (1990). Treating ordinal scales as interval scales: An attempt to resolve the controversy. Nursing Research, 39, 121–123.
Kovač, D., & Trček, D. (2009). Qualitative trust modeling in SOA. Journal of Systems Architecture, 55, 255–263.
Littlestone, N., & Warmuth, M. K. (1989). The weighted majority algorithm. In 30th Annual symposium on foundations of computer science.
Liu, S., Yu, H., Miao, C., & Kot, A. C. (2013). A fuzzy logic based reputation model against unfair ratings. In Proceedings of the 2013 international conference on autonomous agents and multi-agent systems.
Marsh, S. P. (1994). Formalising trust as a computational concept. Ph.D. thesis, University of Stirling.
Marsh, S. P., & Dibben, M. R. (2005). Trust, untrust, distrust and mistrust—An exploration of the dark(er) side. In Trust management. Berlin: Springer.
Pinyol, I., & Sabater-Mir, J. (2011). Computational trust and reputation models for open multi-agent systems: A review. Artificial Intelligence Review, 40, 1–25.
Ramchurn, S. D., Huynh, D., & Jennings, N. R. (2004). Trust in multi-agent systems. Knowledge Engineering Review, 19, 1–25.
Ramchurn, S. D., Sierra, C., Godo, L., & Jennings, N. R. (2004). Devising a trust model for multi-agent interactions using confidence and reputation. Applied Artificial Intelligence, 18, 833–852.
Sabater, J., Paolucci, M., & Conte, R. (2006). Repage: Reputation and image among limited autonomous partners. Journal of Artificial Societies and Social Simulation, 9, 3.
Sabater, J., & Sierra, C. (2005). Review on computational trust and reputation models. Artificial Intelligence Review, 24, 33–60.
Sabater-Mir, J., & Paolucci, M. (2007). On representation and aggregation of social evaluations in computational trust and reputation models. International Journal of Approximate Reasoning, 46, 458–483.
Song, S., Hwang, K., Zhou, R., & Kwok, Y. K. (2005). Trusted p2p transactions with fuzzy reputation aggregation. Internet Computing IEEE, 9, 24–34.
Stevens, S. S. (1946). On the theory of scales of measurement. Science, 103, 677–680.
Taveter, K., Du, H., & Huhns, M. N. (2012). Engineering societal information systems by agent-oriented modeling. Journal of Ambient Intelligence and Smart Environments, 4, 227–252.
Teacy, W. T. L., Patel, J., Jennings, N. R., & Luck, M. (2006). Travos: Trust and reputation in the context of inaccurate information sources. Autonomous Agents and Multi-Agent Systems, 12, 183–198.
Trček, D. (2004). Towards trust management standardization. Computer Standards & Interfaces, 26, 543–548.
Trček, D. (2009). A formal apparatus for modeling trust in computing environments. Mathematical and Computer Modelling, 49, 226–233.
Trček, D. (2010). Ergonomic trust management in pervasive computing environments—Qualitative assessment dynamics. In 5th International conference on pervasive computing and applications (ICPCA).
Trček, D. (2011). Trust management methodologies for the web. In Proceedings of the 7th international conference on reasoning web: Semantic technologies for the web of data.
Trček, D. (2012). An integrative architecture for a sensor-supported trust management system. Sensors, 12, 10774–10787.
Viswanath, B., Mondal, M., Clement, A., Druschel, P., Gummadi, K. P., Mislove, A., et al. (2012). Exploring the design space of social network-based Sybil defenses. In 2012 Fourth international conference on communication systems and networks (COMSNETS).
Weisstein, E. W. (2013). Statistical median. From MathWorld—A Wolfram Web resource. Accessed 21 February, 2013, from http://mathworld.wolfram.com/StatisticalMedian.html.
Yager, R. (1995). An approach to ordinal decision making. International Journal of Approximate Reasoning, 12, 237–261.
Yu, B., Singh, M. P., & Sycara, K. (2004). Developing trust in large-scale peer-to-peer systems. In IEEE first symposium on multi-agent security and survivability.
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
We first introduce terms that we use in the construction.
Lemma 1
Let \(\mathbf {p}=[f_{D}, f_{\textit{PD}}, f_{I}, f_{PT}, f_{T}]\) be a distribution vector. Then
-
(a)
\(0 \le f_\omega \le 1\) for all \(\omega \in \Omega \),
-
(b)
\(\sum _{\omega \in \Omega } f_\omega = 1\).
Proof
Both properties follow directly from Definition 2. \(\square \)
Since the distance metric in Eq. (1) includes a nested expression, which is cumbersome to manipulate, we introduce the notion of cumulative distribution vectors.
Definition 6
Let \(\mathbf {p}=[f_{D}, f_{\textit{PD}}, f_{I}, f_{\textit{PT}}, f_{T}]\) be a distribution vector. Let a cumulative distribution vector \(\mathbf {P} = [F_D, F_{\textit{PD}}, F_I, F_{\textit{PT}}, F_T]\) be computed by successively summing up components of \(\mathbf {p}\) as follows.
We denote distribution vectors and their components with a lower-case letter, such as \(\mathbf {p}\) and \(f_\textit{PT}\), while we write cumulative distribution vectors and their components with upper-case letters, for instance \(\mathbf {P}\) and \(F_\textit{PT}\).
Lemma 2
Let \(\mathbf {P} = [F_D, F_{\textit{PD}}, F_I, F_{PT}, F_T]\) be a cumulative distribution vector. Then
-
(a)
\(F_D \le F_{\textit{PD}} \le F_I \le F_{PT} \le F_T\),
-
(b)
\(0 \le F_D\),
-
(c)
\(F_T = 1\).
Proof
All properties follow directly from previously defined terms.
-
(a)
Follows directly from Definition 6.
-
(b)
Follows directly from Definition 6 and Lemma 1.
-
(c)
Follows directly from Definition 6 and Lemma 1. \(\square \)
We rewrite the metric from Eq. (1) to accommodate cumulative distribution vectors.
Lemma 3
Let \(\mathbf {x}\) and \(\mathbf {y}\) be two distribution vectors, and let \(\mathbf {X}\) and \(\mathbf {Y}\) their cumulative versions as described in Definition 6. Then
Proof
The equality follows directly from definitions of metrics in Eqs. (1) and (2) and the definition of cumulative distribution vectors in Definition 6. \(\square \)
1.1 Equality with median
First, we define the median of a collection. “The median of a collection of values can be found by arranging all the observations from lowest value to highest value and picking the middle one. If there is an even number of observations, then there is no single middle value; the median is then usually defined to be the mean of the two middle values” [42]. But since we operate with qualitative values, we cannot compute the mean. In such cases we pick the value that is the highest between the two. In canonical form, we define the \({{\mathrm{Median}}}\) of a cumulative distribution vector \(\mathbf {P}\) as in Eq. (3).
Second, we compute representative cumulative distribution vectors from representative distribution vectors given in Theorem 1: \(\mathbf {R}_{D} = [1,1,1,1,1]\), \(\mathbf {R}_{\textit{PD}} = [0,1,1,1,1]\), \(\mathbf {R}_{I} = [0,0,1,1,1]\), \(\mathbf {R}_{PT} = [0,0,0,1,1]\), \(\mathbf {R}_{T} = [0,0,0,0,1]\).
Finally, we proceed with the proof. The construction is straightforward: given a cumulative distribution vector, \(\mathbf {P}\), for which the \({{\mathrm{Median}}}\) returns a certain value, \(\omega ={{\mathrm{Median}}}(\mathbf {P})\), we show that the cumulative representative distribution vector that belongs to the same value, \(\mathbf {R}_\omega \), is the closest to \(\mathbf {P}\).
Proof
Let us have a distribution vector \(\mathbf {p}=[f_{D}, f_{\textit{PD}}, f_{I}, f_{PT}, f_{T}]\) and its cumulative version \(\mathbf {P} = [F_D, F_{\textit{PD}}, F_I, F_{PT}, F_T]\). First, we compute the distance between the cumulative distribution vector \(\mathbf {P}\) and all cumulative representative distribution vectors.
The first case we check is \(D = {{\mathrm{Median}}}(\mathbf {P})\). Here, we have to prove the following.
We use expressions from Eqs. (4)–(8), apply basic arithmetic operations and the Eq. (9) simplifies to the following.
Because of Lemma 2, the first component on the right side is the strictest, thus we can drop the remaining three. We obtain the desired result.
The second case covers \(PD = {{\mathrm{Median}}}(\mathbf {P})\). Now, we have to show the following.
We use the greater-or-equal-to operator on the first component on the right side of Eq. (12), because in cases when \(D(\mathbf {P}, \mathbf {R}_D) = D(\mathbf {P}, \mathbf {R}_{\textit{PD}})\), the aggregation method selects PD (the highest between \(D\) and PD; see Definition 4, where we use the maximal operator). We plug in Eqs. (4)–(8) and the expression simplifies.
Because of Lemma 2, the second component on the right side is stricter than third and forth, thus we can omit them. We obtain the desired result.
In the third case, we cover \(I = {{\mathrm{Median}}}(\mathbf {P})\). We show that the following holds.
Again, plugging in Eqs. (4)–(8) yields Eq. (16).
Lemma 2 ensures that the second component on the right side of Eq. (16) binds stricter than the first, while the third component binds stricter than the forth. Omitting the first and forth thus leads to the following.
Now, we check \(PT = {{\mathrm{Median}}}(\mathbf {P})\). We show the following.
Substituting in Eqs. (4)–(8) yields Eq. (19).
Using Lemma 2 we simplify Eq. (19).
Finally, we verify \(T = {{\mathrm{Median}}}(\mathbf {P})\) by proving the following.
Using Eqs. (4)–(8) we obtain Eq. (22).
Because of Lemma 2 we can omit the first three components on the right side, which yields Eq. (23).
We have shown that equivalences in Eqs. (9), (12), (15), (18), (21) hold. This thus proves the equality between the \({{\mathrm{Median}}}\) and \({{\mathrm{Aggregate}}}\) method when using representative distribution vectors from Theorem 1. \(\square \)
1.2 Equality with mean-of-ranks
With the mean-of-ranks we refer to the procedure that computes an average value from a collection of ordinal values by assigning each ordinal value a number, in this case its corresponding rank, and then computing the mean of the collection with the assigned numbers. We give the formula to compute the mean-of-ranks in Eq. (24); we compute it for a given distribution vector. The square brackets denote the nearest integer function, while \({{\mathrm{\textsc {Rank}}}}^{-1}\) stands for inverse mapping from ranks into qualitative values; the mappings are given in Definition 1.
Analogously, we write the mean-of-ranks of a cumulative distribution vector.
Lemma 4
Let \(\mathbf {p}\) be a distribution vector, and let \(\mathbf {P}\) be its cumulative version.
Proof
The equality follows directly from definitions of mean-of-ranks in Eqs. (24) and (25) and the definition of cumulative distribution vectors in Definition 6. \(\square \)
Let us rewrite \({{\mathrm{MOR}}}(\mathbf {P})\) of a cumulative distribution vector \(\mathbf {P} = [F_D, F_{\textit{PD}}, F_I, F_{PT}, F_T]\) in canonical form, so it will be easier to analyze.
Next, we compute cumulative representative distribution vectors from representative distribution vectors given in Theorem 2: \(\mathbf {R}_{D} = [1,1,1,1,1]\), \(\mathbf {R}_{\textit{PD}} = \left[ \frac{3}{4},\frac{3}{4},\frac{3}{4},\frac{3}{4},1\right] \), \(\mathbf {R}_{I} = \left[ \frac{1}{2},\frac{1}{2},\frac{1}{2},\frac{1}{2},1\right] \), \(\mathbf {R}_{PT} = \left[ \frac{1}{4},\frac{1}{4},\frac{1}{4},\frac{1}{4},1\right] \), \(\mathbf {R}_{T} = [0,0,0,0,1]\).
Finally, we are ready to proceed with the proof. The construction is the same as with median: given a cumulative distribution vector, \(\mathbf {P}\), for which the \({{\mathrm{MOR}}}\) returns a certain value, \(\omega ={{\mathrm{MOR}}}(\mathbf {P})\), we prove that the cumulative representative distribution vector that belongs to that value, \(\mathbf {R}_\omega \), is the closest to \(\mathbf {P}\).
Proof
Let us have a distribution vector \(\mathbf {p}=[f_{D}, f_{\textit{PD}}, f_{I}, f_{PT}, f_{T}]\) and its cumulative version \(\mathbf {P} = [F_D, F_{\textit{PD}}, F_I, F_{PT}, F_T]\). First, we compute the distance between the cumulative distribution vector \(\mathbf {P}\) and all cumulative representative distribution vectors.
We write down the distances between a cumulative distribution vector \(\mathbf {P}\) and a all cumulative representative distribution vectors
The first case we check is \(T = {{\mathrm{MOR}}}(\mathbf {P})\). We show that the following holds true.
Using Eqs. (27)–(31) we obtain the following.
Because the last expression in Eq. (33) is the strictest, we can omit the remaining three. Moreover, because of Lemma 2, case (b), we can add the lower-bound.
Next, we address the case of \(PT = {{\mathrm{MOR}}}(\mathbf {P})\).
Next, the case of \(I = {{\mathrm{MOR}}}(\mathbf {P})\).
And the case of \({ PD} = {{\mathrm{MOR}}}(\mathbf {P})\).
Finally, we address the case of \(D = {{\mathrm{MOR}}}(\mathbf {P})\).
Because the first expression in Eq. (45) is the strictest, we can omit the remaining three. Additionally, Lemma 2 states that any component of a cumulative distribution vector is at most 1. Because of this, we can say that the sum of any four components of cumulative distribution vector is at most 4. Therefore we can safely add the upper-bound.
We have shown that equivalences in Eqs. (32), (35), (38), (41), (44) hold. This thus proves Theorem 2. \(\square \)
Rights and permissions
About this article
Cite this article
Jelenc, D., Trček, D. Qualitative trust model with a configurable method to aggregate ordinal data. Auton Agent Multi-Agent Syst 28, 805–835 (2014). https://doi.org/10.1007/s10458-013-9239-8
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10458-013-9239-8