Exploring Digital Diplomacy

In recent works, scholars and diplomats have begun to experiment with ChatGPT, the generative AI bot that has taken the world by storm. For scholars of diplomacy, ChatGPT is yet another digital innovation that may disrupt the work of diplomats. Just like social media, big data, messaging apps and bots, ChatGPT may prove both useful and detrimental to diplomats and diplomatic institutions.

However, ChatGPT is also a unique stage in the digitalization of society. This is because ChatGPT enables regular individuals to interact with AI systems, an activity that was once the purview of programmers who possessed certain technological skills. Moreover, the phenomenal rise in ChatGPT’s popularity, which is growing at a faster pace than TikTok, suggests that similar chat bots may soon become fixtures of daily life. Users may employ ChatGPT to author reports, improve their resumes, write speeches and even grade students’ essays. Most importantly, ChatGPT may soon become an internet gatekeeper. As users employ the chatbot to write jokes, they may also use it to learn about the world asking ChatGPT if it’s safe to visit China, why Russia invaded Ukraine, what values underpin US foreign policy or who discovered America and whether America really land on the moon. As a recent article in the Economist argues, ChatGPT and it’s like may soon replace Google as the main vehicle for searching for information online.  

Over the next few weeks, I will publish a series of posts focusing on the promises, risks and opportunities that ChatGPT presents to diplomats. This week I will focus on one important promise- generating diplomatic texts.

On the one hand, ChatGPT may save diplomats time and energy by generating a wide range of official texts. For instance, UN Ambassadors may ask the chat bot to generate a two-minute speech denouncing a terror attack in Iraq; Embassy staffers may ask ChatGPT to author a tweet celebrating international beer drinking day; negotiators may generate a one page agreement outlining a compromise on important issues while press attaches may create short press statements within seconds.

On the other hand, language matters in diplomacy and an important question is whether ChatGPT is sensitive enough to the nuances of diplomatic terminology. For example, the US may be “dismayed” by the Israeli government’s decision to reform the country’s judicial system. Or, the US may “express concern” over the reform. It may also “condemn” or even “strongly condemn” the actions of the current government. The difference in terminology is crucial as each phrase signals a different intention and may lead to different consequences. If the US is “dismayed”, then the Israeli government may push ahead with its reform assuming that the US does not plan to intervene in an internal crisis. But, if the US “strongly condemns” this reform then the government may pause its plans fearing a crisis with Israel’s strongest and staunchest ally.

To explore ChatGPT’s ability to create diplomatic texts I asked it to formulate a two-minute speech by an Israeli diplomat denouncing anti-Israel bias at the UN. Israeli diplomats often argue that the UN is biased against their country holding Israel to a different standard than other countries. Others claim that Israel warrants increased scrutiny given its policies towards the Palestinians.  I asked for a two-minute speech as diplomats at the UN Human Rights council, often accused of an Israeli bias, can only speak for two-minutes. ChatGPT offered the following speech:

Ladies and gentlemen, esteemed delegates,

I rise today with great sadness and concern, as we once again witness the deeply ingrained anti-Israel bias within the halls of the United Nations. It is truly disheartening to see the international community continue to be misled and manipulated by those who seek to delegitimize and demonize Israel at every turn.

We have seen this bias in the countless one-sided resolutions that have been brought before this body, which fail to hold accountable those who perpetrate violence and terror against Israel, while placing the sole blame for the conflict on the Jewish state. We have seen this bias in the disproportionate attention given to Israel by certain UN bodies, while ignoring the far more egregious violations of human rights that occur around the world. We have seen this bias in the constant attempts to isolate and boycott Israel, which serve only to deepen the divides between our peoples and hinder the prospects for peace.

Let me be clear: Israel seeks nothing more than to live in peace and security alongside its neighbors. We have repeatedly extended our hand in peace, and have made countless sacrifices for the sake of a better future. Yet time and again, we have been met with rejection and violence.

It is time for the international community to wake up and recognize the true nature of the conflict. It is not a result of settlements, or occupation, or any of the other smokescreens that are often used to deflect attention from the real issue at hand. The conflict is a result of the refusal of certain actors to accept the legitimacy of a Jewish state in the Middle East.

We call on the UN and its member states to finally put an end to the anti-Israel bias that has been allowed to fester for far too long. We call on you to stand with us in the pursuit of peace, and to recognize that a true resolution to the conflict can only be achieved through direct negotiations between the parties, without preconditions or ultimatums.

We remain committed to the path of peace, and we look forward to the day when Israel can take its rightful place among the family of nations, free from the constant attacks and baseless accusations that we have endured for far too long.

Thank you.

A veteran Israeli diplomat, who served as the country’s Ambassador to the UN, reviewed the text and said that “it follows the general pattern of Israeli speeches at the UN. Many diplomats, especially younger diplomats, would use this speech without hesitation. But I wouldn’t”. When asked why, the diplomat replied “because it’s too generic. It’s been said too many times. It would generate little reaction in the UN and not get picked up by any newspapers. But if you change some of the arguments and use different phrases you could get your message across and have it picked up by the media”.

Next, I asked ChatGPT to author a short statement by a US diplomat denouncing a terror attack in Israel.  It generated the following text:

“The United States strongly condemns the heinous and cowardly terror attack that took place in Israel. Our thoughts and prayers are with the victims and their families during this difficult time. We stand in solidarity with Israel in the face of this senseless violence and remain committed to working with our allies and partners to combat terrorism in all its forms. Such reprehensible acts of terror only strengthen our resolve to promote peace and security in the region”.

At first glance this text seems identical to similar US Statements published on Twitter or Facebook. However, it is important to note the language used including the strong terms “heinous” and “cowardly”. Moreover, the context is interesting as the statement ends by linking the supposed terror attack with absence of peace, even though I did not tell ChatGPT who perpetrated the attack. It automatically assumed that this attack was linked to the Israeli-Arab or Israeli-Palestinian conflict, a reasonable deduction yet a deduction, nonetheless. Finally, the text ends with a policy announcement- the US will increase its efforts to promote peace.

I then asked ChatGPT to generate a similar statement by a UK diplomat denouncing a terror attack in Israel. Below is the formulated text.

“The United Kingdom condemns in the strongest possible terms the despicable terror attack that took place in Israel. Our hearts go out to the innocent victims and their families who have been affected by this tragic event. We stand firmly with Israel against all forms of terrorism and violence. We call for the perpetrators of this cowardly attack to be brought to justice swiftly. The United Kingdom remains committed to working with our international partners to combat the scourge of terrorism and to promote peace and stability in the region”.

As can be seen, the two texts are similar in from but with subtle changes in nuance. While the “United States strongly condemns”, the “United Kingdom condemns in the strongest possible terms”. America called the attack “heinous” while Britain used the term “despicable”. While America offered “thoughts and prayers”, the UK’s “hearts go out to the innocent victims”. Both statements end with a similar policy orientation- to combat terrorism and bring peace to the region. And yet, there is some difference in tone and language as the UK will “combat the scourge of terrorism” while working with “international partners”, phrases that have in fact been used in many UK statements and tweets.

Finally, I asked ChatGPT to generate a similar statement by a Qatari diplomat, shown below.

“Qatar condemns the terrorist attack that took place in Israel and extends its condolences to the families of the victims. We reject all forms of violence and terrorism, which have no place in any society. We stand in solidarity with Israel in its efforts to combat terrorism and promote peace and stability in the region. Qatar believes in the importance of dialogue and cooperation between all parties to reach a peaceful resolution to the conflict. We urge all parties to exercise restraint and avoid any actions that could escalate the situation further”.

The tone of the final text is quite different. Qatar “condemns” that attack but does not describe it as “heinous”. It “rejects all forms of violence” alluding to the fact that both sides in the conflict use violence. Unlike the American and British statements, the Qatari text emphasizes the “importance of dialogue and cooperation” and “urges all parties” to “exercise restraint” and avoid further “escalation”, again alluding to the fact that both sides can escalate tensions. There is no policy orientation nor promise to promote peace. ChatGPT used a different template, different language and different elements when formulating the Qatari response, as one might expect given the difference in ties between Israel and each nation.

The results of this little experiment offer three important conclusions. First, ChatGPT is already capable of producing diplomatic texts that might serve as a starting point for diplomats. Rather than face a blank page, diplomats can use ChatGPT to generate an initial draft that will then be “fine-tuned” saving time and effort. As AI bots become more sophisticated and are trained using larger and more diverse datasets, it’s likely that generic diplomatic texts will increase in quality. Second, ChatGPT is not tone deaf and seems to tailor diplomatic texts to national language, terminology and euphemism replacing America’s “thoughts and prayers” with the UK’s “hearts go out”. Finally, while the UK and US statements were based on similar templates, a very different template was used in the Qatari statement suggesting the ability to tailor texts based on ties between states. The potential for using ChatGPT to generate diplomatic texts, from comments on recent events to speeches at the UN is thus substantial.

Next week I shall be posting about a different aspect of ChatGPT- its depiction of different countries’ policies and history. An important topic if users come to rely on ChatGPT to learn about world actors and their actions on the world stage.