An excerpt from a discussion on Gamification concepts into Data Labelling Tools and Services
Cogito Tech
Turning Data Challenges into AI Solutions for Enterprises with Precision Labeling, Ethical Sourcing & Global Expertise.
Gamification is a concept very few people know about and fewer know its applications and impact on mundane and repetitive tasks. It is essentially the use of game elements (e.g., points, badges, leaderboards) in non-game contexts. The goal of gamification is usually to motivate people and get them to perform certain behaviors. Researchers have already seen this working quite well in for example fitness apps.
With this objective in mind, Cogito collaborated with Karlsruhe Institute of Technology (KIT) and hosted a chat session between Msc. Research, Mr. Simon Warsinsky working under the guidance of Prof. Dr. Ali Sunyaev (Karlsruhe Institute of Technology (KIT) and our Medical Dentistry expert Dr. Paritrat Prakash Sinha, Project Manager and Functional Consultant, Cogito Tech LLC, to discuss the applications of gamification concepts in data labeling and Medical annotation tasks.
Expert annotation tasks such as medical image annotation are tedious and exhausting but require high attention and engagement from the annotators to ensure high annotation quality. At Cogito, we have always maintained social impact to human resource and imbibed a culture of providing comfortable and causal ambiance to employees to work without any stress. Thus incorporating gamification into the data labeling job is something we explored.
For labelers, gamification could make the labeling process more engaging and thus make them more motivated and ultimately increase the quality of the provided labels. Some examples for gamifying labeling tasks would be to give labelers points for successfully completing tasks or including a competition. However, what is usually not easy to understand is, how gamification has to be designed in a specific context so that it actually works.
The 9 key principle identified by Karlsruhe Institute of Technology (KIT), and feedbacks by Dr. Sinha would allow annotators with some of the following benefits:
-???????provide annotators with easy to comprehend feedback about their annotation performance
-???????communicate the purpose of achieving the desired outcomes of the system
-???????include gamification elements that allow annotators to interact with each other and form a sense of community at their own discretion.
-???????provide annotators with optional purely hedonic activities that are detached from annotation
-???????include gamification elements that represent both short-term, performance-based goals as well as long term, effort-based goals, and associate each goal with a reward
The core problem to resolve was to make a monotonous and repetitive task engaging in a clever way. The question is straightforward: How can we get the users to enjoy this monotonous and repetitive task? Multiple different game concepts based on design principles were thought out and in the end their socio and psychological impact are observed. Our expert showcased his observations on the Medical Data Annotations and impact of similar rewards systems which were implemented here at Cogito.
The discussion concluded several similar outcomes and feedback from our expert project manager which shall impact data labeling jobs and also would be considered in the development of data annotation tools and software. The data labeling tools market is going to grow at a CAGR of 30% between 2021-27, and possess great opportunity to incorporate gamification concepts.
As a future work, we need to investigate how to incorporate and integrate these gaming elements while development of data labeling tools and observing its impact on the quality of the data annotations and interest levels of the data annotators. With an efficient game to build the annotations, the game/tool creators could try to offer lower monetary reward than the other competitors. It is important to note that the game would need to be constructed in an efficient manner, in a way that as little as possible answers are needed to build the annotations.