5 Q’s for Zafer Elcik, Co-Founder of Otsimo
The Center for Data Innovation spoke with Zafer Elcik, co-founder of Otsimo, a company that develops education games for children with autism, based in Ankara, Turkey. Elcik discussed how gamification is useful for teaching children with autism and how machine learning can deliver appropriate educational content.
This interview has been edited for clarity.
Joshua New: Can you describe why you decided to develop software for children with autism?
Zafer Elcik: My brother has autism. Despite his inability to talk, read, and write, he has a special interest in smart devices. Seeing him spending time on my phone for up to 55 minutes was a blessing. However there were no games in the market that directly focuses on children with autism or are developed for them. Some support apps are available in the market, but they are quite expensive—some of them cost $250. However, Otsimo been able to reach over 20,000 kids with autism.
New: How does Otsimo stand out from other kinds of treatments for autism?
Elcik: Otsimo is a mobile educational platform for children with learning disabilities, especially autism. It includes educational games, offers reports to parents, and optimizes itself based on its analysis of game data, such as true-false rates, gestures, session time, and heat map analysis. Otsimo games are created through the “Applied Behavior Analysis” (ABA) method. ABA is the usage of certain educational techniques to implement positive and meaningful changes in specific behaviors of a person. One of the techniques is positive reinforcement, which we use in our application. The educational aspect of ABA games is a proven solution for learning disorders.
New: How is gamification particularly useful for helping children with autism?
Elcik: Education efforts towards an individual with autism is only effective when it holds their attention. In that manner, the longer you draw their attention, the more effective the education process for that the child becomes.
However gamification for autism is different. Many children with autism have a problem with the Theory of Mind concept, which is the ability to perceive how other individuals feel, think, and behave. Because of this we couldn’t create a gamification system on empathy and understanding. To solve this problem, we create a child-centric system that relies on positive reinforcement. Our gamification rewards the child for the process, not for the best answers. Trying is bigger than knowing!
New: How effective is this approach at reducing autism symptoms?
Elcik: Our approach is effective for most cases. We have hundreds of pieces of great feedback and dozens of children who have learned how to read and write with our application. We relied on a large research base to achieve this. We collaborated with special education educators, psychologist, and several autism foundations in Turkey to develop our games. Researches based on ABA teaching techniques on tablets and smart devices concluded that children with autism and related disorders can improve their performance by utilizing these devices. The result correlates with our feedback and we are quite proud for that!
Average attention span for children with autism is around 5 to 15 minutes. But with our apps we have data that shows this attention span can increase to 30 to 50 minutes after using our application.
New: What kind of data does Otsimo collect as children play games? What can this tell you?
Elcik: We have more than 12 million anonymized playing data points. We create 2D feature sets with 37 features representing the actions of a particular child playing a game. In other words, these can be considered pools of specific event values of a child.
The most important data points are the game score and positions and timings for specific events. Thanks to that data we can analyze the child performance on single items as well as entire categories of tasks. We use data to find the best difficulty and in-game experience and tweak the entire platform to meet the children’s needs.
The system uses machine learning to understand the “repetition cycle” of the child and adjust difficulty accordingly. For example, if some values are considered low for that specific difficulty, the difficulty is downgraded. In that manner, games become less challenging and hints are displayed more frequently. Games can also get more difficult if a child consistently does well. This allows us to put the right content in front of children at the right time.