Trust-Aware Elastic Social Compute Units

Abstract

Abstract-With the advance of research in human computation, applications and software systems are increasingly being designed to include the human aspect of computation. We work with Social Compute Units (SCUs) that are computational constructs with people as their core resources who use software services to organize and conduct their work. SCUs are collaborative units, and have a cloud-like behavior in the sense that they are elastically provisioned and adapted at runtime. Systems that utilize the concept of SCUs bring challenges that are associated with the highly dynamic and unpredictable human-centric behavior. Thus, trust in human based services is of paramount importance. While there is related work on social trust in the social networking and crowdsourcing areas, trust in highly coordinated constructs such as SCUs remains a significant challenge. In this paper we provide a trust model that considers merging social trust with performance-based trust of human based services into an integrated trust model for SCUs. We illustrate the models' application in concrete strategies, such as for elastic management of SCUs and incentives for SCU members.

Application of provenance in social computing: A case study

“…The most important metrics to mention here are effort , productivity , willingness , reliability , and performance trust , which is a weighted aggregate metric of the aforementioned ones; social trust , which we define as a weighted aggregate of votes from people that the worker has collaborated with; and socio‐technical trust , which is a weighted average mean of the aforementioned performance‐based trust and social‐trust metrics. For the definitions of the aforementioned metrics, we refer you to a previous work of some of the authors of this work, presented in the work of Riveni et al We modeled SCUs as lists of ICUs, which, in turn, have their own metrics. A task is designed with specific properties, such as the skill type required to execute it, the cost for executing it, and the deadline to execute it.…”

“…For the definitions of the aforementioned metrics, we refer you to a previous work of some of the authors of this work, presented in the work of Riveni et al 18 We modeled SCUs as lists of ICUs, which, in turn, have their own metrics. A task is designed with specific properties, such as the skill type required to execute it, the cost for executing it, and the deadline to execute it.…”

“…Depending on the domain, an SCU provisioning platform may have its proprietary pool of resources, but human‐based resources/services can also be invoked from multiple external sources, such as from crowdsourcing platforms, enterprise pools of human resources, social networks, and others. We have implemented a proof‐of‐concept prototype with ranking, monitoring, collective‐adaptation algorithms, and profile updates, presented in our previous work, and here, we present our updated prototype, which includes mapping of our collective‐adaptation algorithm result‐logs to PROV terms and adding provenance‐data to our prototype, including visualization of ICU, SCU, and task‐related provenance data.…”