AI: Floodlight Instead of Flashlight in the Cavern of Data

 |   Microsoft News Center

DPMK

Did you know that there is a correlation between the number of early school-leavers and the number of classes taught by head teachers? Well, the people who operate the school system didn’t know it either, until now. This is one of the things advanced analytics can point out.

 All school systems try to keep early school-leavers to a minimum. (Officially early school-leavers are young people aged 18-24 who do not have secondary school qualifications and are not receiving education.) “The Hungarian rate is around 12-13 percent at the moment, the problem with this is not that it is a little higher than the EU average, but that it has been rising for some years,” says Ádám Horváth, head of the Digital Teaching Methodology Center (DTMC) at the Microsoft Future Decoded conference. We need to understand the causes of this process to be able to reverse it as quickly and effectively as possible.

The DTMC thinks that the problem will be solved by adopting AI-based analytics tools, and luckily, practically all the conditions for this already exist. The most important is perhaps data: the Education Office collects an outstanding amount of high quality data about almost every single element in the education system, which is exceptional even on the European level. However, something was still missing: the tools to simply and quickly analyse all this data. The use of the previous tools required specialised knowledge, and the analyses could not be run on a large amount of data within a reasonable amount of time. However, Ádám Horváth said that the use of cloud-based AI tools offers numerous advantages. The processes and decisions may be based on factual data, the information can be extracted significantly faster and with less manpower, and as a result of self-service systems, decision-makers gain direct access to the information.

dpmk2The DTMC has already completed the first pilot project. The purpose of this was to check the methodology, to determine how suitable the selected method is for performing fast data analysis, so they only used institute-level data. All the available data was loaded into the system to reveal any concealed correlations without setting up any hypotheses or making any assumptions in advance. The huge amount of data was the very reason that such comprehensive analyses could not be performed manually. “In the past it was like searching for something in a dark cavern with a flashlight, we only ever saw one small point clearly. Now though we are able to light up the entire cavern, and obtain a comprehensive picture of the whole situation,” said the head of the DTMC highlighting the difference.

However, we need to be clear about the limitations of such data analysis systems as well: they can identify the correlations, but they are unable to interpret them. They can’t tell what is the cause and what is the effect, and under no circumstances are they able to decide what kind of intervention is required. This remains the work of the education specialists, who may do this much easier after they are given the pertinent information.

One unexpected correlation that was revealed was that the school-leaving ratio is significantly higher in schools where the head teacher is also a class tutor and teaches many classes, and where there is no school secretary who is able to take on some of the administrative burden. Of course, discovering a correlation such as this does not in itself describe the problem, but it does call attention to it. Planning the intervention depends on innumerable other factors too, but at least we know what factors actually exist. Later on, the data analysis processes that are created can also be used to examine correlations between anonymous person-level data.

Predictive analytics is a special data analysis method that “learns” the relationships among the data available and creates data that characterise future states. The greatest potential advantage of this method is that models created in this way assist decision-making and selecting the path towards the desired objective.

AI-based analytics tools are also capable of the real-time analysis of big data; through this the information created from the data can be directly and quickly returned to the institutions as feedback—even at classroom level.

AI-based analytics tools have proved their worth in numerous fields. Even the first trials are showing that they can be successfully used in education too, thereby providing effective help for decision-makers.

YouTube Video

For Hungarian version please click the link below: 

https://news.microsoft.com/hu-hu/2018/11/28/mi-zseblampa-helyett-reflektor-az-adatok-barlangjaban/