"Grengjai" in Different Contexts: From Meeting Other Cultures to Conversations with AI

The concept of Grengjai

The concept of "grengjai" is very well-known and sometimes even subconscious to Thais, but something which has never failed to puzzle westerners when they learn of it. The concept revolves around the desire not to cause discomfort, inconvenience, or burden to others, often at personal cost. 

The internet is filled with foreigners asking about this concept on online forums like Reddit and social media posts. I was confused. Why is this concept so difficult for them to understand? 

My experiences

What I understood when I moved to the UK is that it's not that foreigners lack the concept of being respectful or considerate of others; rather, the difference lies in the nuances in the balance and expectation. They would rather be straightforward about their wants, but still expect the other to prioritise their own interests if the request inconveniences them. An alternative approach to being considerate and deferential.
I found that grengjai depends on mutual recognition. Without reciprocation, what was meant as consideration can instead become quiet (and needless) self-sacrifice. Here in the UK, I find it crucial to stand up for my own interests in class or interviews, trusting that others will do the same. 

Learning when to withhold grengjai and when to voice my needs has become part of adapting to living abroad. It doesn't mean I'm throwing away my cultural values, but simply choosing the time that's appropriate to use it.

How about AI?

As someone working extensively with AI systems, I began noticing something curious: I was unconsciously applying grengjai even when prompting language models. I often found myself typing phrases like "Could you please help me with…" or "If it's not too much trouble…" to these models, which made me realise that the concept of grengjai is so ingrained in our culture and way of thinking that it naturally bleeds into the way we talk and ask for favours, AI included.

This begs the question, does our naturally "grengjai" way of writing and asking for favours affect how well we get responses from AI (particularly GPTs)? This question is especially noteworthy when it comes to prompting these GPTs in another language, as it would have learnt primarily from non-Thais and thus not treat grengjai language as the norm.

The importance of this lies in the fact that common GPT models generate words or groups of words (known as tokens) based on what is most likely to come next, given the text that comes before (including the prompt). Non-thinking models don't interpret the semantics of the question, but instead treat the text itself as the input. So whilst "Can you please explain how GPTs work?" and "Explain how GPTs work" mean the same to us, they're treated as entirely different questions by GPTs.

So does it affect the results? Yes, it does due to the aforementioned principles of GPTs, but whether it affects how well the AI responds to our prompts is much harder to tell and is worth a whole academic paper in and of itself. But what's clear is that there's an environmental impact. Sam Altman, the CEO of OpenAI, has claimed that they lose tens of millions in electricity costs from users adding "please" or "thank you" out of politeness. This is because AI models process text token by token and thus a longer prompt generally requires more computational resources. A grengjai-style prompt unnecessarily adds to the prompt length.

This, of course, comes with environmental impact, as these small politeness add up to real energy use, exacerbating the industry's already massive carbon footprint, with services like ChatGPT alone generating over 260,000 kg of CO₂ monthly.

What this means for me

I'm not saying "grengjai" is bad or counterproductive, but just like everything else in life, it has its place. It's important to understand your situation and know when to apply it. Now with the prominence of AI, it's no longer just for yourself; it's for the future and the environment, too.

Story by Pakkapol Lailert