We present you around 20 statements on politics (varying between municipalities) and after you have answered these, we present you with the candidate that have opinions most similar to yours. Now you can explore This way voters could see what candidate or party were making the best match. To cover the elections in 98 municipalities and 5 regions we collected data from more than 7000 candidates. Apart from guiding voters in learning about the candidates and whats at stake in the election we also provided autogenerated texts about the results for every single of the 10515 candidates based on election
The voting advice applications have documented impact on voting behavior, and Altingets tool have helped more than 500.000 voters (15 pct of the electorate) to decide how to cast their vote. The application both contain the answers of the candidates on a row of political statements (on which the result is calculated) and some background questions about the candidate fx campaign topics, personal description, and other personal information that the candidate can choose to share. This way, voters have a unique way to get to know the candidates – or use this information to make up their choice.
The data we collected from more than 7.000 candidates was available for the newsroom for journalism. Many conclusions could be made on politics based on this large survey of political actors’ opinions. A row of articles about all from schools, environment, sports, health, and culture were published during the runup to the election based on how the candidates distributed their answers in the survey for the application.
We also made an automated text generation tool that publishes a personal online article on the election result for every single candidate. As soon as the votes were counted and calculated we could publish no less than 10515 articles in a single moment to give both candidates and readers a better insight in how the votes had impact. The articles were connected to the candidates’ profile in the voting advice application and were sent to every candidate by email. The candidates themselves shared their article on social media.
We built the voting advice application in our own system containing database, API, backend, survey tool and frontend. In the system we also had the calculation method/algorithm to make the most fair matching between candidate and voter.
We also took public available data about the election result and used it to generate personal articles for evey candidate. We used Python and a system of if-sentences.
What was the hardest part of this project?
Many aspects of this project were complex: – It was a challenge to develop the right calculation method. Many critics point at the risk of bias according to what algorithm you use as well as how you choose your answering ouptions. After extensive tests and analysis, we came up with the “5-minus-1 method”, that we deem as the most fair calculation. – It was also a challenge to choose and distribute the right statments in the application. The choice of statements is really sensitive as it is very easy to formulate in a biased way or let out important topics. Therefore we have been working on the statements in an editorial process for one year including methodology, research, formulation, tests on candidates, on usersegments, on experts and finally a large internal test and analysis to make the sheet of statements balanced and relevant and with adequate discripancy between parties, so we can present a useful result for the voters. Technically we developed a system for the statements to be more adaptive to the municipalities. That means a devision of statement classes. 1. ideological statments goes for all (also acros regional and municipal elections) 2. basic questions goes for all municipalities (or all reggions). 3. we made five groupings of municpalities according to size and location. We made statements for these specific groups. 4. Specific statments for each municipality or region made the application more relevant around the country. this meant more than 300 statements in total. It was a challenge to make sure all Finaly we introduces a word merger to substitute “my municipality” with the name of the municipality in each statement to give more local feel. – It was also a major challenge to collect all relevant info from the more than 7000 candidates as well as profile
What can others learn from this project?
One learning is that methodology is very important. Since the tool have effect on the election results, it is important to remain unbiased and that statements and calculation have high quality.
Another leatning is that it takes time to prepare and especially during complex/multiple elections as municipal ones, it is important to start working on the allpication early on.
But this type of project can lead to both great journalism and high (sometimes extremely high) traffic numbers.