Motivation and goal:

Covid 19 being the first global modern day pandemic, it has influenced our society in many ways since it started 4 years ago. Causing many deaths and forcing our society to impose heavy restrictions, we wanted to observe if it had any effects on beliefs, such as philosophy and religion.

We will observe any changes in beliefs mainly by looking at the changes in views for relevant subjects.

Research questions:

1. Do Wikipedia’s philosophy and religion page views provide insights into how individuals perceive life amid the unprecedented social conditions of COVID-19?

2. How do these perceptions vary across countries, and do cultural differences in main cultural philosophical currents influence philosophical and religious shifts?

3. Can we identify groups of cultures exhibiting similar philosophical and religious reactions to the changes induced by COVID-19?

Dataset description:

We use an API to retrieve the views from the philosophy pages. The name of the philosophies are taken from the english page list of philosophy. Then we get the url for each english philosophy for all languages and construct our dataframes. We use the daily granularity to have enought data points for short period of time, in particular between the first case and the lockdown. When we have the dataframe for all languages, we keep only the philosophy pages that exist in all target languages to have a more uniform comparison.

In examining philosophical themes’ attention shifts on Wikipedia during the COVID era, it became evident that the overall trends in philosophical topics mirrored the broader global trends. Consequently, delving into specific movements within philosophical ideas or clusters proved more compelling than focusing on philosophy’s general trends as a whole.

Part 1: Looking at how people interests across philosophical topics behave during Covid 19 period.

To analize different trends within the views of the different english wikipedia pages, we have formulated 5 broader philosophical topics based on information obtained from Wikipedia pages, supplemented by insights from ChatGPT, to group similar topics together.

On the pie chart above, we can see how total page logs are distributed across our constructed broader philosophy topics, political and religious related ideas are obviously a large chunk of total views while we see that more theoritical philosophy current are less represented.

To begin the analysis, we will gather aggregated time series data across five relevant periods: the period before COVID (prior to the first case), the period from the first case to the start of lockdown measures, the period during lockdown, and the period after COVID. By adding up the page view values for each philosophy page within each broader topic, we can create time series plots to identify any intriguing trends. One unique observation is for religion. It is the only topic not experiencing a significant decrease in interest during the Christmas period! This observation is not entirely surprising.

Now let’s investigate what’s happening at crucial covid-19 dates (delimited by vertical dashed lines).

First covid case date (red line):

We could expect a rise in pageviews due to an initial period of uncertainity and information seeking. However, this chart doesn’t show a clear spike in pageviews around this time, which could suggest that the initial spread of COVID-19 may not have immediately influenced interest in philosophical topics, due it it being slow, far away, or deemed unimportant, for example.

Lockdown Date (blue line):

The lockdown period is typically associated with people having more time at home, potentially leading to increased internet usage. In this case, we can see that some topics like Metaphysics and Political Thought show a marked increase in pageviews shortly after the lockdown begins. This may indicate that during periods of enforced isolation, either people had more time to engage with complex subjects, or sought philosophical perspectives to make sense of the societal changes occurring around them or both.

Back to Normalcy Date (green line):

As restrictions ease and a sense of normalcy returns, we could expect the interest in philosophical topics to decrease as people return to their regular activities. However, the chart shows that for some topics like Political Thought and classical philosophy, the interest remains steady, even rising after the return to normalcy, which might suggest that the pandemic had a lasting impact on people’s engagement with these subjects. It’s important to consider that these interpretations are speculative, and the actual reasons behind the changes in pageviews could have multiple sources.

We will now try to look at the following correlation heatmaps that displays correlation between time series of page views for every topic on every period considered to hopefully understand better dynamics between topics.

Before COVID-19:

• High positive correlations are seen among several philosophical clusters, indicating that people who are interested in one philosophical topic are often interested in others as well.

First Case to Lockdown:

• During the initial phase of the pandemic, there is an overall increase in the correlations between topics, suggesting that the onset of the pandemic led to a more unified interest across different philosophical areas.
• Science philosophies now show strong correlation with other topics, perhaps due to the increased interest in scientific understanding of the pandemic and its implications.

During Lockdown:

• The correlations between topics in this period are generally lower compared to the period from the first case to the lockdown, which might suggest that the topics diverged in terms of pageview behaviors during the lockdown.
• Science philosophies show a significant correlation with Ethics and Morality that was low before covid, which could be due to discussions about the ethical dimensions of scientific research, public health measures, and the role of science in guiding policy during the crisis.
• Eastern and Western Philosophies appear to have a decreased correlation with other philosophical discussions, possibly indicating that during lockdown, cultural or regional philosophical interests became more concentrated within their own spheres or less influenced by global philosophical trends. Correlation between the 2 as also been divided by two compared to pre-pandemic period, showing some growth in regional interests.

After Lockdown:

• Globally we see decreased correlations compared to before covid and even lockdown period reflecting a diversification of interests.
• We see that compared to before covid, correlation between ethics and morality and science as more than doubled, it jumped up during lockdown and remained at this level potentially showing persistent change in how people perceive and link science with ethical considerations as a consequence of the intense ethical debates that characterized the height of the pandemic.
• In essence, lasting changes would be evident if the post-pandemic correlations do not revert to the levels seen in the "Before Covid” heatmap. This would indicate that the pandemic has had an enduring effect on the way people engage with and relate different philosophical topics. A return to the "Before Covid" correlation levels would suggest that the pandemic's impact was temporary and that philosophical interests have normalized as the immediate crisis has receded.

Having looked at data in a somewhat qualitative manner, we will now try to dive into some more rigorous and causal approach for digging further into this question with the next part of our data story where we will adopt a more statistical rigorous approach.

Part 2 : Variation of trends for philosophical pages in English

The goal is to retrieve the trend and see if the timepoints linked to Covid-19 have a significant impact on this trend.

The Data preprocessing :

The study now is made on the dataframe containing the pageviews for all philosophy in english. The granularity of the views are daily and it can be seen on this graph for example, that it makes a lot of noise, so the data is converted to have weekly points. As shown here, the noise is less visible.

Another preprocessing step is the selection of significant philosophies. Indeed, the disparity of number of logs per page is very large. Although it has already been filtered by using only pages that exist in all the target languages, this disparity might impact the relevance of the study. To prevent that, a threshold of 10% of the most viewed page is taken, and only pages above are kept for the rest of the analysis. This leaves us with 68 philosophies.

First observations on the trends :

The study of this trend can not be done directly on this raw data of pageviews. Indeed, the potential change in interest for a certain philosophy is only interesting if it is caused by the particular situation of the pandemic, and not an unrelated variation of interest. To mitigate this issue, the seasonality of the interest is separated from the trend. Thus the usual variations that occur over the year for a certain page are not supposed to impact our analysis.

Change in the trends :

Using the principle of interrupted time series, the trend was segmented into 3 time periods. Before Covid-19, during and after, these periods were separated by two timepoints : February 16th 2020 and June 21st 2021. These dates were chosen to match the average dates of the first case and end return to normalcy. This part of the study uses linear regression, so the dates inside the Covid period were not selected because it gave segments that were too short to properly study. On each segment, a linear regression was made. Here is an example with the Metaphysics page, that had a significant change at t0 but not t1.

What can be observed is that even if the slope is not changed significantly, sometimes there is a big difference in the baseline, i.e. in the intercept and that is why we also study it. This phenomenon can be seen in the example with Marxism, at t0.

These plots display the variations in the slope of the trend and in the intercept. In other words, it shows the difference between what was predicted from the previous tendency and what actually occurred. It allows us to have a better view over significant changes for the philosophies, and choose more accurately which groups to look at. Now let’s look at these graphs precisely and more closely to the meaning of these philosophical currents.

• Challenging the political system

  When the covid arrives, among the philosophies that have a strong increase in their interest are Socialism and Populism. This is interesting because they are both political philosophies that are driven by social inequalities and that criticize the capitalist system. Close to these currents are Communism and Marxism, that have a strong increase in the predicted intercept. Also, we can see that although Fascism has a slight slow down in its trend, it remains a highly popular page and after the beginning of covid has an increase in the predicted intercept. Finally, pages Anarchy and Anarchism also have higher baselines after the beginning of Covid. After the Covid, the tendency is reversed for Socialism but the others keep their positive trends. This is interesting because during the Covid, inequalities were highlighted more than ever, especially during lockdowns. Depending on social class, differences in quality of life, education and financial security have been exacerbated more than ever. This could explain the renewed interest, particularly in populism, which had previously been on the wane. What’s more, the measures taken by governments have often been criticized and judged to be very harsh, which could also explain the stronger interest in philosophies linked to the power of the state and the questioning of its system. Notice that we can observe a slight drop of views at the end of the graph but that is common in a lot of pages at this moment, probably due to the full return to normalcy leading to decreasing interest in wikipedia in general.

• Religious questioning

  For the Baptist (and Baptism) we can see a slight acceleration in the trend, then a slowdown again, but overall the change is not very marked.

  For Agnosticism, we can see that interest increases before Covid, stagnates, decreases slightly during and drops drastically afterwards. What’s interesting here is that the variations are very marked between the first and last periods. This philosophy is rooted in not knowing. It’s interesting that after this period of confinement, which was very conducive to reflection, views dropped like that. On the contrary, the beginning of the trend for Atheism**, is comparable to the one for agnosticism, with a more steep decline during Covid, but the after Covid has a very clear increase in interest.

  For currents like Paganism or Pantheism, that take root mostly in the connection we have with the world, especially the natural world, the trend drops after return to normalcy and during lockdown as well. Especially for Paganism that had rising interest before. It seems plausible that being deprived of contact with the world and nature, the beliefs that rely on this connection can take a hit.

  Finally New Age and Mysticism are both trending before and after Covid, but not during. These currents are more spiritual than religious and revolve around personal growth for one, and experiencing communion with a divine reality for the other. The lack of connection with the world, the narrowness of our lifestyle during this period might have made it hard to pursue these beliefs.

Part 3: Looking at how languages dynamics across philosophical articles behave during Covid 19 period.

Now, we want to observe how the dynamics of the consultation of the philosophy related pages evolved before, during and after Covid across different languages. In other words, we analyze how the consumption of wikipedia philosophy related pages is different across languages. The first thing that we can take a look at is the total views of philosophy pages in each language in log scale to reduce discrepancies between languages. We just want to take a look at the trend.

In the analysis of the graph, a distinct trend emerges as languages collectively ascend during the Covid period. The upward movement suggests a shared trajectory, possibly indicating a global response to the challenges posed by the pandemic. Notably, Japan exhibits a delay in its upward shift, which could be attributed to its implementation of lockdown measures at a later date compared to other regions. This temporal discrepancy in response strategies is mirrored in the graph. Intriguingly, certain languages appear relatively unaffected, maintaining a consistent course, while others display persistent changes, underscoring the diverse impacts of the pandemic on linguistic patterns.There is a general trend of increased page views around the first vertical line, which could correspond with the onset of Covid where all countries show a peak in interest. The second peak corresponds to the onset of lockdowns, where the interest in philosophical content increases or maintains a high level. The decline in page views after the second peak may indicate a gradual return to normalcy or a saturation point in the public's search for philosophical content. Next, we decided to analyze correlation between languages in order to observe which languages move together and which languages move differently. Therefore we compute correlation matrices for each period.

In order to make these results clearer, we plot these dendrograms to show how different language editions of philosophy-related Wikipedia pages clustered together before, during, and after the COVID-19 outbreak based on the patterns of pageviews.

• Before COVID-19: There is a distinct cluster that groups the Scandinavian languages (Danish, Norwegian, Swedish), suggesting similar trends in philosophy pageviews. The Romance languages (Italian, French) and Germanic languages (Dutch, German) form separate clusters, with Japanese and Korean being outliers.
• During COVID-19: The cluster patterns shift, showing a Scandinavian group remaining intact, while Italian joins the German and Japanese cluster. French and Serbian form a distinct cluster, separating from Dutch.
• After COVID-19: The clustering shows further divergence, with Finnish and Korean forming a separate cluster, and Japanese becoming an outlier. The Scandinavian languages no longer form a tight cluster, indicating a possible shift in the shared attention of philosophical topics.

These changes in clustering patterns over time could reflect cultural shifts in attention to philosophical discourse during different phases of the pandemic. So, here we have a first clustering of our data based on trends of views in each language. We showed that certain languages can have a comparable consumption of philosophy on wikipedia. But, the lack of this analysis is that they can consult totally different pages. Therefore, in order to have a detailed analysis we will study this aspect. For that, we computed a fraction of views of each article at all time steps in each language. And we tried to repeat the same analysis as before, try to observe if some languages have a similar consumption of philosophy pages on wikipedia.

This heatmap represents the dissimilarity in pageview patterns between different languages. Lower distances suggest more similar patterns.

• The heatmap illustrates that Danish and Dutch, as well as German and French, have closer viewing patterns, indicated by the lighter colors.
• On the other hand, Japanese and Korean show larger distances from other languages, suggesting unique patterns of engagement with philosophical content.

In order to visualize this more clearly, we reduce our data’s dimension so that we can plot all data points in two dimensions.

• Clusters in this scatter plot further illustrate the relationships between different language editions. Languages that are closer together in this plot have more similar pageview patterns.
• Korean and Japanese are close in this plot, which aligns with the post-Covid dendrogram, suggesting a similarity in the shift of attention in philosophy-related content.

And finally, we performed a clustering analysis of our fraction of views to effectively determine if clusters exist or not. For this analysis we choose two methods. The first one uses the fraction computed before, during and after covid. The second one, we tried to reduce the noise in our fraction. We subtracted the mean of fraction of views before covid to make differences appear more clearly. Each dot representing a cluster.

Data processing	Before Covid	During Covid	After Covid
Raw fraction of views	French, Italian, Serbian Japanese, Korean German, Swedish, Danish, Finnish, Dutch, Norwegian	French, Italian, Serbian,German, Swedish, Danish, Finnish, Dutch, Norwegian Japanese, Korean	French, Italian, Serbian Japanese, Korean German, Swedish, Danish, Finnish, Dutch, Norwegian
De-noisy fraction of views	French, Italian, Serbian Japanese, Korean German, Swedish, Danish, Finnish, Dutch, Norwegian	French, Italian, Serbian,German, Swedish, Danish, Finnish, Dutch, Norwegian Japanese, Korean	French, Italian, Serbian,German, Swedish, Danish, Finnish, Dutch, Norwegian Japanese, Korean

We found that clusters built with a reduction of noise matches the cluster determined with correlation in dendrograms. However, this clustering using fractions of views is less clear than the one using correlation. We observe a pretty similar distribution on fraction of views on each article in each language except for Korean and Japanese. We can clearly say that after all this analysis when Covid came the consumption of Wikipedia philosopy pages has never been more similar accross languages. In summary, the analysis suggests that there were distinct shifts in the consumption of philosophical content across different language editions of Wikipedia during the Covid pandemic. Cultural and linguistic similarities influenced these patterns before the pandemic, which shifted as the pandemic unfolded and persisted even after the peak of the crisis. These shifts may reflect broader changes in cultural and societal attitudes towards philosophical topics during times of global crisis.