Insights

Artificial Intelligence in Japan: The Path Forward

Key contacts

Kaitlin Outtrim

Project Manager & Researcher

5 years of experience unlocking research insights for commercial and government clients across a variety of sectors, with a focus on bridging ties between Japan and global markets

Overview

In our last AI article, we discussed how Japanese consumers are cautious about the positives that AI innovation can bring, and are nearing fatigue. So where to from here? We’ll discuss how to best leverage AI in a cautious market like Japan.

Where We Left Off 

In part one of our AI series, we focused on the current “State of the Nation” for AI use and sentiment in Japan, which remains cautious compared to other markets. In our second and final part, we discuss the path forward and real world implications for future AI technologies, including localisations recommendations for new market entrants. As always, we draw on our first-hand experience working with clients and consumers here at CarterIGNITE to share with you the key insights for success in Japan. 

Japanese Language and AI 

Japanese is not an easy language, especially for English speakers. AI systems, particularly large language models (LLMs), accelerate the quality of DIY translations, but the technology is still not completely accurate, and therefore not well-trusted by Japanese users. 

CarterIGNITE recently worked with a corporate LLM client to assist their localisation from Western to East Asian markets, covering both Korea and Japan. Compared to research findings in Korea, AI models working in the Japanese language had a much wider room for improvement due to the added complexity of the Japanese writing systems and the subtleties of workplace interactions.  

As the majority of LLMs are trained via English language resources, whose outputs are then machine translated into Japanese, common errors uncovered in our research interviews include; 

  1. Misunderstanding word meanings due to errors in kana/writing system interpretation (i.e., deliberately using the Japanese language’s three writing systems in unconventional ways for stylistic purposes). 

  2. Sounding “too American” when prompted to write advertising copy or social media posts. 

  3. Not “reading the air” (what English speakers call “reading the room”) and failing to understand the nuance behind what people say, rather than relying on literal or direct translations, creates misunderstandings. 

The Japanese language relies heavily on context, as speakers often deliberately leave key information unsaid to ease communication or avoid a confrontational tone. As LLMs “learn” from English-based materials that lack this cultural layer of understanding, our research participants frequently mentioned that they cannot completely rely on Japanese language LLM outputs for the accuracy of information or the quality of translation.  

Therefore, if you are contacting Japan-based businesses or potential partners, we highly recommend using human translation and/or interpretation for all communication. This is particularly important in formal business scenarios; in our own experience in translating English to Japanese, Google Translate, DeepL, or ChatGPT can only capture some of the nuances required. Japanese people prefer a “more is more” approach to communication, so it is better to lead with as many accurately translated materials (such as company descriptions and brochures) as possible.  

Replacements for Human Interaction? Not so Likely   

Japan is the world’s test market for AgeTech innovations. After all, it’s a well-known fact that we are home to the oldest population on the planet, and this statistic will only grow in the years to come. CarterIGNITE and our sister company CarterJMRN are no strangers to robotics in AgeTech and aged care research, including the use of AI tools inside care facilities to plug the gaps of a stretched-out (and shrinking) Japanese labour force.  

However, our research has found that the current performance level of AI-powered support tools is insufficient to replace or assist humans in all instances. We recently collaborated with a global robotics company that developed an AI-powered robot designed to engage in conversation with care facility residents, aiming to increase social engagement, particularly for individuals with Alzheimer's or Dementia-like symptoms.  

Our observations and interviews with residents who trialled the Japan-localised model found the robot difficult to engage with at times, as it asked questions that can feel invasive or inappropriate for surface-level small talk in Japan (for example, asking details about late family members) or tended to switch between dialects.  

A further study of an in-home helper prototype robot further reinforces the insight that people, particularly older adults, prefer human interaction to machine interaction. When posed with the possibility of completely customising an in-home assistance robot that could perform the tasks of a human carer (such as lifting from beds and chairs, assisting with exercise and physical wellbeing, and social interaction), our Japanese research participants voiced their preference for a human helper for a surprising reason: the lack of social conflict with a robot.  

The older adults we spoke to value the spontaneity of human engagement, as they expressed a disinterest in a conversation partner that passively agrees and obeys everything said, as AI tools currently do. For a bonus regional insight, this desire was strongest among participants from the Kansai region, an area known for being more direct and outspoken compared to counterparts elsewhere in Japan.  

Market Research and AI  

Market researchers play a critical role in consumer marketing and product development, as we act as interpreters for what people say and what they actually mean. As explained in the earlier coverage of Japanese language and AI, a significant portion of this work requires human engagement to ensure the accuracy of the interpretation of findings, data analysis and reporting.  

Our industry is no stranger to the innovations that technology can bring. It is remarkable to imagine how challenging quantitative data collection once was when relying solely on street interviews and telephone fieldwork. While these methods remain valuable tools, online surveys now allow us to gather thousands of responses over a weekend rather than over weeks or months. 

Current discussions with industry organisations such as ESOMAR reinforce that our core functions should remain human-operated, with AI used as an efficiency tool only. Data hallucinations and incorrect interpretations during analysis and reporting stages are incredibly common when relying on AI tools (such as LLMs). Compared to quantitative methods, qualitative methods remain relatively hands-on.  

AI-assisted technologies, such as transcription and note-taking services, can add efficiencies that allow researchers to focus on deeper-level thinking, insight extraction, and connecting with our research participants in the cultural context. Researchers discuss replacing our human participants with bots or “synthetic data,” and we have identified this as a growing quality control issue in quantitative research panels. However, since true insights arise from new ways of thinking that LLMs and other AI tools cannot replicate, we fail to see the benefits beyond creating echo chambers of readily accessible information.     

The Next Steps in AI 

So where does the current phase in the “AI Era” leave us? We certainly have the ability to accelerate our productivities in some areas, but we must remain vigiliant of its limitations and (current) high error rate. Environmental damage caused by excessive power and water consumption in AI generation is also a cause for concern.  

Looking forward, we believe the duty lies on individual businesses and creatives to consume and create with AI tools in ethical and responsible manners; we cannot rely on laws and regulations to set the guidelines for us.  

AI tools certainly enable market researchers to fill the gaps in “quick and dirty” style research projects with budget and/or timing restrictions (of course with prior client approval), but no one can replicate the creativity of human thought. 

Imitations and copyright concerns are top of mind for Japanese consumers. This subject is brought up frequently in our AI research, particularly in relation to AI image and video generation which replicates Miyazaki Hayao’s Studio Ghibli art style. Art and artisans continue to contribute to Japan’s rich history, culture and pride. Japanese consumers value the work and training that goes in to honing and perfecting a craft, again, something that AI cannot replicate.    

As mentioned, the nuances of the Japanese language compared to English (the default language in which majority of AI tools train) currently make it difficult for humans to rely entirely on content that AI generates.  

Japanese older adults in particular value human-to-human connection to keep their minds sharp and to bring variety into their routines. Meanwhile, younger consumers aged under 20 in Japan are the most knowledgeable of the benefits that AI can bring, but remain apprehensive of its implications in fueling misinformation and jeopardising their job security.  

AI can assist in product localisation and translation (e.g., machine translations on first draft – depending on the complexity of the original text), but bilingual speakers must review final versions to avoid errors.   

And finally, as Japanese consumers remain fairly unenthused by the wave of AI-enabled products and services hitting the market, ensure that any AI functions and features genuinely offer a point-of-difference that makes sense for the market.  

So What?  

To sum up the points discussed across our two articles; Japan is certainly ready for AI technologies, but only if they provide a meaningful improvement to quality of life or efficiency. In a market where high quality products gain the most trust, market entrants must be selective of the ways that they engage in AI tools to avoid negative consumer reputation.   

Your Partners in Japan Market Entry & Research 

Do you have more questions about your market success in Japan? Contact our team at CarterIGNITE for more information or a brief introductory call.  

We are a team of experienced market researchers and market entry specialists who assist businesses and governments at every stage of their journey into Japan. CarterIGNITE employs a range of research methods, including qualitative and quantitative approaches, as well as consultancy services, in addition to trade mission and business matchmaking support across various industries. We’d be delighted to support you and “ignite” your entry into Japan and beyond.