Research

Human Centric Systems and Social Media

 

Human Centric Systems and Social Media group aims to analyse, understand human behavior and optimise the flow of information in a social network to make organizations function effectively.

Vision
Collective intelligence of humans is the most critical asset of an enterprise. Social Media is expected to make enterprises, smarter, social and more productive by leveraging collective intellectual with effective knowledge sharing and engagement strategies. For successful social media adoption, user participation is vital to reap the benefits expected from social platforms. Our research is focused on incentives design, sentiment analysis, improving the quantity and quality of knowledge contribution on social platforms towards extracting more value from them. We will look at algorithms and non-intrusive user interfaces to improve human engagement, build a semantic knowledge base, simple levers for crowdsourced problem solving and social learning.

Research Topics

Our core interest areas include:

  • Social Network Analysis
  • Gamification
  • Crowdsourcing
  • Behavior analysis for personalization 
  • Semantic Web–Linked Data

We are taking a deep dive into Information Quality and Diffusion, Nature of incentives and motivation patterns, Game mechanics, Reputation Graph to locate experts, Semantic Taxonomy and Progressive Learning Projects.

These are the must do things to stay focused on our vision to design human centric systems in a way to leverage human capital & social media at large:

  • Modeling social reputation to locate subject matter experts, key influencers and peer groups
  • Recommendation systems that doesn’t suffer from filter bubbles
  • Crowdsourced compute engine for distributed problem solving
  • Building dynamic learning flows for progressive learning
  • Separating temporal content from lasting knowledge in UGC platforms to increase the longevity  and user retention in a network
  • How various aspects of privacy affect participation and growth?
  • Studying nature of Incentives & Motivation to trigger right game mechanics
  • Analysing the quality and diffusion nature of information
  • Social thermodynamics as an experimental approach for network analysis
  • Understanding human bias, nature of feedbacks and churn in social platforms
  • Leveraging social proof and ontological commitments for game dynamics and realizing system goals

Game Mechanics and Human Behavior
Gamification is intended to derive a behaviour in response to a stimuli. Our focus is to implement various applicable game-dynamics model into the organisational social-ecosystem and analyse the nature of motivational outcomes. Human motivation is categorized as intrinsic and extrinsic based on whether its an innate activity or a reward driven act. Intrinsic motivations are difficult to induce but are more powerful than extrinsic motivation. We largely focus on developing extrinsic motivation, which over constant and beneficial exposure may transform to intrinsic motivation. We are aiming to address various issues relating to user adoption, user retention and personalisation of the platform. The core focus will be on creation of models for personalisation, behavioural analysis, behaviour tagging and user engagement. One of the major test beds we are using is Knome.

Social Network Analysis
A rich functional enterprise social graph would bridge the gap between our primary goal of optimizing the information flow and problems in achieving that. This information and network analysis tool will provide us the capability to ease the data collection and problem solving.  The graph structure aims to capture and culminate all the important cross-domain relations (how people are connected to each other, how a particular topic is connected to other topics and users, affinities between a person<-> person, person<-> content, content <-> content) into a graph data structure, to serve as a strategic data hub that serves contextual relations on demand. It will have the ability to weave into all platforms and services [internal and external] with a common data header, to let data flow into our system seamlessly aiming to expose unnoticed relations between people, content and interests.

Information Quality
This is probably the most crucial aspect for the survival of a social network or any UGC platform, yet gets minimal attention in research. Evaluating credibility of content and people is the first step in the direction to improve quality. Main parameters are depth of information, completeness of content in terms of references backed by factual information, validity. Separating temporal content from lasting knowledge on social platforms improves the longevity of network and also enables us to provide better contextual results. The problem is more striking in enterprise where networks intend to be both knowledge sharing and general networking.

Understanding Bias and Feedbacks
Biases of people who help information spread inadvertently affect information. By accumulating these false biases over time, we are looking at a totally false or twisted version of story. Information credibility is at stake here, so are the decisions built on this. Discovering various types of existing cognitive biases in social networks is the aim. As part of this we will be looking into identify conflicting opinions (first step for pushing self opinion), removing common misconceptions, classifying authorized and unauthorized sources, cross comparing with sentiment in online reviews and ratings to arrive at more accurate information.

Semantic Taxonomy for Enterprise
Easier discovery of content is of vital importance. Can we make a library of tags available for content? How can enterprises do this, in deep relation to the content they use? While users are usually tag providers, can the system help them with a ready to use set? How semantically rich and related can we make this taxonomy? How can we grow this and link this across platforms in a fairly automated way, so that the big picture emerges for the business leaders.

Progressive Learning
Enterprise learning platforms are often cast in iron with very little room for individual choices on modes of learning, type of courses preferred. These platforms often ignore the tacit knowledge of employees and go by tests or courses taken that may not deliver as much value as hands on experience and knowledge. Besides, traditional learning platforms may not incorporate new media and content freely available on the internet. Also enterprises rarely consider multidisciplinary learning, which is getting to be crucial. We are looking at customizable learning paths and diversity in learning. This approach will   take into account the distance among similar sets to promote maximum diversity in the learning flows. This will leverage our earlier work, (the feedback framework) designed to judge diverse facets of a personality from interactions on the web.

The group is headed by Ashok Krish

 

Reach Us.

Share