Trust and context in user-centred data ecosystems

The following is excerpted from the May 2014 World Economic Forum report Rethinking Personal Data: Trust and Context in User-Centred Data Ecosystems. To read the full report, visit weforum.org/personaldata.  

Despite the growing recognition of the importance for understanding the context of data usage to inform effective policies, little work has been done in this regard. To address this concern, a collaborative global research initiative was established between the World Economic Forum and Microsoft. This appendix is a summarized version of the full report published in May 2014.

The intent of the study was to examine how individuals define context, focusing on the factors that impact their sensitivity regarding the use of data related to them by service providers. The project studied how these factors vary across different countries, how they can aid in the design of context-aware system, and how these systems can be integrated into user-experience designs for interactions that are more meaningful and consistent with complex individual preferences.  

The results show that a variety of factors, both objective and subjective, influence the perception that individuals hold regarding the appropriateness of a given scenario. Other demographic characteristics unique to each individual – for example, their age or level of technological sophistication – also play a role. 

Figure 16: Factors impacting individuals’ sensitivity to the use of their data

 

Source: World Economic Forum

Framework for analysis

Throughout 2012 and 2013, Microsoft sponsored a series of research studies to address these issues. The research was divided into two stages. The first phase involved qualitative studies in Canada, China, Germany and the United States to develop insights on users’ mental models on their personal data. The second phase provided quantitative analysis to validate the initial insights. For this phase, the original list of countries was expanded to include Australia, India, Sweden and the United Kingdom. The research identified seven distinct factors that individuals consider when determining whether a given use of data is acceptable. This is defined as the data context.

The importance of including subjective variables underscores that data use context is very much defined by personal preferences. Note that neither “trust” nor “value exchange” on its own is sufficient to determine acceptability of data use to the individual. They play major roles, but they do not pre-empt other factors. 

In addition to these seven variables, factors related to the mental models of individuals are also identified. Some of these individually-based factors included:

• Attitudes to and adeptness with technology
• Awareness of the relationships and activities within the personal data ecosystem
• Perceptions of government protection

The impact of contextual factors

In most geographies, collection method had the largest impact of all the variables examined. Similar to other research demonstrating that individuals want to have a sense of control over how data is collected, it is interesting to note the strength of this desire despite results that show inconsistent behaviour, perhaps due to the relative lack of available tools for individuals to effectively manage this attribute. 

The trust variable was the third most important factor determining acceptable use in the Western countries. For situations involving the passive collection of data and where individuals perceived no additional benefit (which causes low acceptability), the trust variable had a significant positive impact for all countries. 

Figure 17: The impact of contextual factors

 

Source: World Economic Forum

Although value exchange had a smaller impact in the Western countries, it had the second largest impact in China. In addition, the research uncovers preliminary evidence that individuals tend to frame their interactions from the perspective of a perceived value exchange. When the value exchange is to deliver benefits to users – either in saving time and/or money or to enable something of unique value – the acceptability rate is highest for all countries. However, attitudes towards value exchange when presented in terms of community benefit were more variable, possibly reflecting differences in cultural values between different countries.

The research results discussed here show that individual preferences for data use are nuanced and contextual. With subjective factors such as trust in service provider, perceived value exchange and other attitudinal, demographics and cultural factors all playing a role, what is considered acceptable is clearly personal and will evolve over time. 

Binary approaches to data governance that treat all data as equal, and apply universally, are thus neither appropriate nor flexible enough, especially in a world of big data. Incorporating context-related nuances into regulations is difficult. However, technologies similar to those described here may provide an alternative, by facilitating policy frameworks that are principle- and outcome-driven, rather than process- or technology-driven. 

Context-based systems and user experiences

A better understanding of individuals’ perceptions of context can contribute to the development of systems that incorporate contextual elements into governance and individual engagement functions. Data governance systems that incorporate contextual elements enable more user-centred data ecosystems by respecting individual preferences in data-use scenarios. Where individual preferences are unknown, recommend personalized settings that are based either on individuals’ past preferences or prevailing practices.

One approach is a “recommender system” that can be deployed either on behalf of service providers to enable a personalized user experience, or on behalf of individuals as “personal assistants” to help with context-sensitive data settings for different types of applications. In either case, these systems minimize the probability of data uses that are inconsistent with user expectations and empower individuals to engage in more meaningful interactions with service providers, thus increasing the level of trust in the overall ecosystem. 

Conceptually, a recommender system that can assess a number of variables can help predict the acceptability of a given data-use scenario, and recommend appropriate data settings, either to an application or a user. If it predicts a negative decision from the user, it can share with the proxy what additional factors would make the scenario more acceptable. A more meaningful user experience can be achieved either by providing additional input or enabling the users to negotiate on the factors that the research found to have the largest impact on increasing acceptability (trust, value exchange, and data usage).

Figure 18: Illustrative example of a recommender system

 

Source: World Economic Forum

Policy implications

Context is key, even if not well understood. The research presented here shows context as being driven by multiple variables – some objective, but some clearly subjective – and driven by continuously evolving social and cultural norms. However, defining such abstract concepts in regulations is not ideal, and can lead to overly prescriptive and less adaptive laws. 

Technologies such as those described here enable the development of context-aware systems, and an alternative approach to policy frameworks that respects individual preferences and needs according to the context of a given data usage. This is different than emphasizing the initial context(s) in which the data was collected. This difference, and the technologies that facilitate it, are crucial for trustworthy data ecosystems. 

Importantly, policy development can be informed by evolving technology and research. As the latter advance, new insights into individual behaviours and preferences, and proof of concept on the technology front can influence the scope and flexibility of regulatory frameworks. Policy-makers can base accountability regimes around outcomes rather than a fixed set of rules.

Over time, context-aware systems can be coupled with other technologies such as a metadata-based architectures – where data is logically accompanied by interoperable “metadata tags”. These tags can contain use policies associated with the data and related provenance information. Combined, these preferences and permissions can inform any entity that touches the data on how it can be used. Providing automated mechanisms that can facilitate contextually appropriate data use can also be leveraged for its enforcement. What is considered acceptable context would be reflected in the data-use policies – examining these policies would reveal contextually inconsistent uses. With these, innovations, it would be easier to uphold principles in a constantly changing world of big data. 

More research is needed on how context can be defined more clearly and simply, and how it can be practically integrated into systems and interface designs that engender meaningful user engagements. This research is needed at the global level to provide an evidence base that can be used to develop an interoperable global framework, or simply a framework that would allow individuals from one region to access services in another – a basic enabler for today’s internet commerce.