The value of trust: understanding the scale and depth of the user’s perspective

Koen WillaertShenja van der Graaf

IT as a Utility ConferenceSouthampton

Starting point

Context

Importance of trust in online environments is widely acknowledged.

Even though trust is important in real world too, this is even more so online due to the lack of the providers’ physical presence and the rare frequency of transactions between two entities.

Trust definitionTrust in an ICT system is a property of individual users of the system representing their subjective view about the system.

Users who don’t trust the system are disinclined to use it.

OPTET takes the view that trust is a good thing, because it helps users to benefit from new, innovative ICT systems which they might otherwise reject.

Trust definition broken up

Two actors: trustor + trustee Risk assessment + decision making Two outcomes: trust level and the trustor’s

decision to engage (or not) with a trustee by taking into account the associated risks and benefits.

… has received sufficient attention, yet general tendency to readily assume that trust is intrinsically beneficial…

Is trust intrinsically beneficial?

‘A market in which participants are trusted to the correct degree is as efficient as a market with complete trustworthiness’ *

Thus, high trust is not panacea but trust should correctly reflect trustworthiness!

* Brainov, Sandholm, “Contracting with uncertain level of trust”, In Proceedings of the 1st ACM conference on Electronic commerce (EC '99)

Finding the correct balance

TrustWorthy

Trust

Information Asymmetry

Goal OPTET: reaching a next level of automation in this process

10

Gap

Social sciences:Trust antecedentsImpact of ‘trust by design’ solutions….

Computer sciences:Computational modelsAutonomous decision makingReputation systems/ recommenders …

Gap

Social sciences:Abstract/general levelPredicting trust in specific system in particular situation?Subgroups/segmentation? Researching Trustworthiness?

Computer sciences:….it is widely recognized that trust is highly subjective… but the effects of trustor’s attributes are in most cases ignored in computational trust models

Goal

Explore for both individual trustors as for organisations/service providers issues related to finding the right balance between Trust and Trustworthiness

Research and integrate subjective attributes in a trust computational model for individual trustors towards an online system

Survey research based on model McKnight

Results Individual trustors

Methodology

Datamining on 8 trust related concepts:

Trust stanceTrust beliefs in general professionals Institution based trust General trust sense levels ICT-domain specific sense of trust levels Motivation to engage in trust related seeking behavior Trust related competencesPerceived importance of trustworthiness design elements

Survey details:

Online survey Period: February-March 2013 28 countriesMajor representation of UK (32%) and Greece (18%)90 subjects included in analysis Mostly 5 point Likert scale itemsK-means clustering for segmentation purposesAnova

16

Trust seeking behaviour

18/03/2014 17

Individual trustors segments

Ambivalent

22%

Ambivalent

22%

Highly Active trust

seeking

31%

Highly Active trust

seeking

31%

Medium Active trust

seeking

20%

Medium Active trust

seeking

20%

High Trust

27%

High Trust

27%

Segment 1: Ambivalent Trust

Ambivalent

22%

Ambivalent

22%

Medium to high trust stance

Moderate seeking activity

Low competence level

‘Forced’ trust Simple heuristics

Segment 2: High Trust

High trust stance

Limited seeking activity

Medium competence level

High Trust

27%

High Trust

27%

Segment 3: Highly active trust seeking

Low to medium trust stance

Very thorough trustworthiness investigations beyond cues from service provider

Medium competence level

Procedures in case of problems

Highly active trust

seeking

31%

Highly active trust

seeking

31%

Segment 4: Medium active trust seeking

Low to medium trust stance

Medium level seeking activity

Medium to high competence level

Procedures in case of problems

Medium Active trust

seeking

20%

Medium Active trust

seeking

20%

Trust levels

Results Organisations / service providers

Methodology Organizations

Datamining on trust related concepts:

Trust stanceTrust beliefs in general professionals Institution based trust General trust sense levels ICT-domain specific sense of trust levels Motivation to engage in trust related seeking behavior Trust related competencesActive assessment of the trustworthiness of own and third party applications Trustworthiness design elements

Survey details:

Online survey Period: February-March 2013 Pool of international, national, regional and local organisations 48 commercial organisation + 57 public organisations Mostly 5 point Likert scale itemsK-means clustering for segmentation purposesAnova

25

Organisational profile segments

Low TrustLow

competence

6%

Low TrustLow

competence

6%

Low Trust High

seeking

28%

Low Trust High

seeking

28%

High TrustHigh

seeking

39%

High TrustHigh

seeking

39%

High TrustLow seeking

26%

High TrustLow seeking

26%

Organisational segment 1: Low trust – Low Competency Lower ability to understand

terms of services and to detect threats and misuses

Trustworthiness assessments are likely to be less efficient or inaccurate

Negative views towards the professionalism and the expertise of other parties

Ambivalence Lower ability to provide

guarantees to customers Lower ability to provide

information about applicable laws governing their relationship with their customers

Low TrustLow

competence

6%

Low TrustLow

competence

6%

Organisational segment 2: High trust – Low seeking Risk of overestimating the

actual trustworthiness Lower motivation to collect

trustworthiness information (for instance real time monitoring)

Intend to provide less guarantees to customers

Intend to provide less information about applicable laws governing their relationship with their customers

Less concerned about the influence of current and near future technology trends on online trust/privacy and online trust/security

High TrustLow seeking

26%

High TrustLow seeking

26%

Organisational segment 3: Low trust – High seeking Less risk inducing behaviors

than the two previous clusters Less positive view on current

and future safeguards, regulations and technological advances

Low Trust High

seeking

28%

Low Trust High

seeking

28%

Organisational segment 4: High trust – High seeking High trust stance and positive

views on current and future safeguards, regulations and technological advances.

Active seeking behavior High TrustHigh

seeking

39%

High TrustHigh

seeking

39%

Implications

Mapping Individual and organisational segments

Low Trust High

seeking

28%

Low Trust High

seeking

28%

High TrustHigh

seeking

39%

High TrustHigh

seeking

39%

Highly active trust

seeking

31%

Highly active trust

seeking

31%

Medium Active trust

seeking

20%

Medium Active trust

seeking

20%

High Trust Low seeking

26%

High Trust Low seeking

26%

Low TrustLow

competence

6%

Low TrustLow

competence

6%

Ambivalent

22%

Ambivalent

22%

High Trust

26%

High Trust

26%

Tailored ‘Trust by design’ solutions

AmbivalentAmbivalent

Referrals Easy to scan,

simple and straightforward cues

Familiarity Technically working

well

Tailored ‘Trust by design’ solutions

Quick scanning of cues

Look & feelHigh TrustLow seekingHigh Trust

Low seeking

Tailored ‘Trust by design’ solutions

Clear terms of service + data policy

Contact information Outline on legal

frameworks Offers a complaint

mechanism Large set + variety

of cues

Highly Active trust

seeking

Highly Active trust

seeking

Tailored ‘Trust by design’ solutions

Clear terms of service + data policy

Display the financing organization

Medium Active trust

seeking

Medium Active trust

seeking

Risks associated to organisational trust profiles

High Trust Low seekingHigh Trust

Low seeking

Low TrustLow

competence

Low TrustLow

competence

Low Trust High

seeking

Low Trust High

seeking

High TrustHigh

seeking

High TrustHigh

seeking

Negative outlook on future trends and developments

? Inaccurate trustworthiness estimation,

Inability to explain terms of services / offer guarantees

A too high trust level in comparison to the actual trustworthiness

Conclusions + outcome

Conclusion

Both congruence + asymmetry between individual trustors and organisational profiles

Diverging implications Prioritizations in terms of actions for

organisations/service providers Services providers need to develop ‘trust by

design’ solutions targeted various trustors segments

Outcome: Trust Modelling and Estimation based on subjective attributes

40

High Trust High Trust Ambivalent trust

Ambivalent trust

Highly active trust seeking Highly active trust seeking

Medium active trust seeking

Medium active trust seeking

Trust stance Trust stance Trust related competences Trust related

competences Motivation to engage in trust seeking behavior Motivation to engage in trust seeking behavior

User segments

Core trust attributes underlying segmentation

Translation of trust attributes into trust estimator

Future work