Talking Afghanistan: Discourse Analysis of Pakistan and US Based Experts on Afghan War


In the backdrop of fall of Kabul on 15th August, G5iO conducted a study to understand the variations in the narrative and analyses given by different Afghan policy experts of Pakistan and the US. In doing so, we have analyzed their writings both on social media and mainstream media. The goal of this exercise is to show the (in)consistency in the statements and narrative on the Afghanistan from both groups of experts. 


The purpose for nuance and balanced analysis we picked five academics and policy experts from each the US and Pakistan who have been writing on Afghanistan. We focused particularly on those with an active presence on Twitter by drawing on their various stances and assessments as presented on their Twitter timelines.

#US Experts (Affiliation)Pakistan Experts (Affiliation)
1Christine Fair (Georgetown University)Muhammad Amir Rana (Pakistan Institute of Peace Studies)
2Elizabeth Threlkeld (Stimson Center)Amna Khan (Institute of Strategic Studies Islamabad)
3Peter Lavoy (Policy Maker)Ejaz Haider (Journalist)
4Joshua White (Brookings Institution)Maleeha Lodhi (Diplomat)
5Michael Kugelman (Wilson Center)Zahid Hussain (Journalist/Policy Analyst)

After finalizing the list, we downloaded their Twitter timelines. According to our collected data the experts from the US tweeted 11,952 times whereas Pakistani scholars tweeted 15,885 times in total. Next, we combined the data and filtered out only those tweets that were in English and focused

only on Afghanistan. We did that by using a string search of keywords (i.e., Afghanistan, Afghan). This resulted in a total of 3232 tweets, 1626 from the US and 1606 from Pakistani experts, thus comprising our final data set for analysis. As for traditional media, we looked for their published online news or op-eds in different national and international newspapers to understand their narrative and policy inputs which subsequently served as additional data informing our discourse analysis.

Social Media Analysis

  • Tweeting behavior

The first thing we looked at was the tweeting frequency behaviour of both groups. The graph below indicates that while both groups have similar behaviour, there are however certain differences. Experts in Pakistan began discussing the Afghanistan issue as early as 2019 when the possible US-Taliban deal was in sight. As such, they remained consistent in their frequency of tweets related to Afghanistan throughout 2019-20. In contrast, US experts remained mostly silent on Afghanistan and only began talking about the issue around after the spring of 2020. Their frequency of tweets however, picked up as the US-withdrawal neared.

  • Narrative (di/con) vergence

In addition to how often they were talking about Afghanistan, we wanted to know what exactly was being said as well. In order to do that, we conducted a correlational analysis of their most frequently used words to know whether both groups were looking at Afghanistan via a similar or differing perceptual lens when it comes to Afghanistan. The following graph shows that both groups talked about the ‘Taliban’, ‘peace process’ and ‘Kabul government’ along similar lines

However, there were some key divergences worth noting. For example, the US experts often talked about ‘US-withdrawal’, ‘America’s plan’ and ‘troops’. Whereas, Pakistani experts appeared to focus more on ‘dialogue’, (the) ‘future’, and their ‘relationship with afghans’.

The red-regression line predicts the likelihood of words used in the tweets by both the US and Pakistani experts. Additionally, words that are further from the red lines represents no relationship in terms of their usage between the two group of experts.

  • Topical analysis:

Next, we wanted to explore their narrative even further and offer a computational thematic analysis. In doing so, we used an unsupervised machine learning technique of topic modelling that uses the Latent Dirichlet Allocation (LDA) algorithm to classify main themes. The topic modelling of tweets from both groups adds further nuance to the analysis.

The following graphs indicate the differing uses and prevalence of specific topics. In the case of the US, one can see that their experts always discussed the US-withdrawal policy which gradually picked up pace as the withdrawal deadline withdrawal came closer. Whereas, references to the peace process and negotiations rarely made an appearance in their tweets.

On the other hand, Pakistani-experts have mostly talked about the ‘afghan people’, the ‘on-going talks’, and the ‘peace process’. In addition to this, Pakistani experts were quite diverse in their tweets where they offered their opinions on a broader range of issues and hence, did not focus on just one specific issue.  

The figure illustrates topic modeling that through mathematical method classify similar words used in tweets and cluster them together. Moreover, one can see in the figure above that tweets by Pakistani experts when talked on Afghanistan can be classified in four different topics and each topic is further represented by four key terms. Similarly, the US experts’ themes are also classified by calculating their usage of key terms into four distinct topics and the prevalence of each theme over a period of time.  

  • Echo-chambers and homophily 

In our study, we wanted to understand whose opinion these experts valued and whether there was a convergence between them on the issue of Afghanistan. In order to do this, we conducted a social network analysis where we took two metrics: mentions, and retweets under-consideration. This analysis offered interesting insights.

The network on the left-hand shows that the US-based experts often mentioned or retweeted each other especially when it came to the Afghanistan issue. Hence, there was a strong inter-connectivity or echo chamber like structure suggesting propagation of a similar narrative.

On the contrary, the Pakistan-based experts often held diverse views on Afghanistan where only two experts (Zahid Hussain and Amina Khan) were found to have often converged on the Afghanistan issue. The rest however tended to be more independent from each other’s views. Hence, their opinion was not propagated within specific communities and instead resonated in a wider audience.

Furthermore, it is interesting to see that the experts in Pakistan continued mentioning/retweeting “US4AfghanPeace” tweets. It is the account of Zalimay Khalilzad who represented the US and struck a deal with the Taliban in May 2020. Interestingly, he is not mentioned or retweeted among the US policy experts who propagated and amplified the views of people who also held opinions similar to their own.

  • Sentiment analysis:

Lastly, we aimed at conducting a sentiment analysis to understand the tone and emotions that were being communicated by each group of experts. The following visualization show clear variations between the sentiments expressed by both groups. For instance, the experts from the US were increasingly more negative as the US-withdrawal came closer. In contrast, the experts from Pakistan talked more positively at least after 2019. As such, their score of negative tweets is also less than that of the US-experts, i.e., -4 and -6.

Download PDF:

Talking Afghanistan: Discourse Analysis of Pakistan and US Based Experts on Afghan War

Share on facebook
Share on twitter
Share on linkedin