Is Cultural Sensitivity a Challenge for Artificial Intelligence?

You cannot have missed it: Artificial Intelligence (AI) is the talk of town. Like with every new tool, it takes some time before all its possibilities are harnessed to their full extent. The technology is most likely here to stay in the field of localization, so why not try to think of ways it could make the process more accurate?

Localization implies a deep understanding of both the source and target languages and of their cultural characteristics. Cultural sensitivity checks are common in the industry, but they sometimes become critical with certain projects, such as marketing texts, video subtitling, training materials, etc.

We thought, “Why not ask ChatGPT what it would do to conduct a cultural sensitivity check?” to see where it stands, and how to make it even better with human input. Let’s go!

Prompt: List the items that ChatGPT would check during a cultural sensitivity check

Here’s a breakdown of some of the answers we got:

Cultural References

• Verify that cultural references in the original text are accurately translated and appropriately adapted for the target audience
• Ensure that the translated content does not include references that may be offensive or misunderstood in the target culture.

Human Input

Cultural references have always played a major role in hooking the reader into a marketing strategy. AI can sometimes identify such references, but will it be able to determine which equivalent is appropriate in the target language? Perhaps a reference to American football in a source text would for instance need to be adapted to soccer in France, but to Gaelic football in Ireland, or to cricket in India. Once that has been determined, an actual, meaningful equivalence has to be found. This implies creativity.

As far as references that should be avoided, these are susceptible to change on a daily basis, according to politics, or news. AI can help make sure we don’t miss any, but human involvement remains essential. Can AI make the decision to keep or remove references to Twin Towers for the US market, or to cartoon caricatures for the French market?

Taboos and Sensitive Topics

• Evaluate whether the translation avoids using language that may be considered taboo or offensive in the target culture.
• Check for potential sensitivities related to religion, politics, gender, or other sensitive topics.

Human Input

Are anglicisms acceptable to use in all marketing campaigns in Quebec? For all topics? French-Canadian people getting trained in big organizations have their own terminology. The Quebec Board of the French Language spends a lot of time and resources updating the languages with new French words that should be used in official contexts. AI can identify terms that might need to be double-checked. We’ll leave the double-checking to the human eye.

On the other side of the Atlantic, in France, Belgium, and Switzerland, the French language has evolved in other ways. Inclusive writing, while not officially recognized by the Academy, is becoming more popular, and might sometimes be preferable for companies targeting a younger, more politically involved audience. But when to use it or not is a matter of opinion, and of client preference. Let AI ask the question, “Should inclusive writing be used here?”. You linguists will know the answer.

Humor and Wit

• Assess if humor and wit in the original text are effectively translated without losing cultural relevance.
• Verify that jokes or references maintain cultural appropriateness and do not unintentionally offend.

Human Input

Now, this is a bone of contention that will not go away anytime soon. AI can identify jokes it already knows and try to find an equivalent if it has been trained on content that includes multilingual humor. Still, the exercise is risky and often disappointing. Humor is already subjective by human standards, so leaving it to a machine will get us nowhere. What of humor based on puns? On image/text discrepancies? What if the source text uses slang, a type of language less present in AI training material? AI will sometimes help identify offensive content, as previously discussed, but the wit still has to come from the heart.

Historical Sensitivity

• Verify that historical events and references are accurately presented and do not perpetuate historical inaccuracies or biases.
• Check for any content that may evoke sensitive historical memories.

Human Input

AI can read history books, but can it read the room? Is this PowerPoint presentation written in English suitable for all French-speaking markets? Former French colonies in Africa might not find equal value in references to some historical French events. Does Christopher Columbus still evoke the same national sentiment in America today, or for every population segment? Can a marketing campaign for the Cantonese market use the same historical references as in the Mandarin-speaking regions?

AI is a powerful tool, and asking the right question will get you closer to the right answer. As such, it can help you establish a complete checklist for cultural sensitivity that linguists or LSPs can use to make sure all texts are perfectly adapted to every locale.