Is AI like a new-generation "The Sims"? Virtual family members are becoming a business

The game industry has been discussing “emotional companionship” for many years, but when a The Sims–like setup moves from virtual communities into real families, the topic is no longer just about gameplay design; it turns into an ethical multiple-choice question.

The controversial case this time involves a mother in her eighties who has been video chatting almost every day for a year with the “son” on her screen—chatting on time, giving reminders on time—yet has no idea that her real son has died in a car accident. To preserve the elderly woman’s will to live, the family turned to Zhang Zewei’s team, who used the son’s lifetime footage, voice, and dialect data to generate a virtual avatar capable of real-time interaction, restoring his appearance and speech habits as faithfully as possible. Objectively speaking, this is more like an extremely vertical “custom NPC system,” except that the target users are no longer players, but bereaved family members.

From a product perspective, the team no longer sees itself as selling AI software, but as providing “emotional tech services,” and its business model is also being adjusted: restoration projects for martyrs are offered for free, while ordinary users pay low renewal fees to maintain the service. This pricing is more like a long-term subscription-based online game, trading a low entry barrier for stable retention. Changes on the demand side are also obvious: orders for emotional restoration are rapidly increasing, indicating that the technology has moved beyond the early “curiosity-driven users” phase and is beginning to touch a broader range of real-world needs—a process similar to a niche indie game gaining a wave of growth after being boosted by social media.

The points of controversy mainly focus on two directions: first, whether virtual companionship will hinder psychological healing, and second, the potential secondary trauma when the truth is revealed. From a design standpoint, this kind of “AI son” is less a companionship system and more a continuously running immersive storyline. It deliberately maintains a worldview in which “the son is still working away from home,” and it offers no ending option. This is the complete opposite of most games, which provide an emotional outlet in the final chapter of the main story. Supporters believe such digital figures can serve as a “refuge,” providing vulnerable groups with short-term emotional support; opponents worry that if one’s life remains stuck in a disguised save file for too long, reality will eventually deliver a much more violent lesson.

Objectively, this project currently looks more like a “trial version of a second-generation product”: the functionality is in place, but the boundaries are far from tested. The team itself emphasizes that the core purpose of the technology is to serve the living and alleviate pain. They deliberately reduce follow-up contact after project delivery to avoid secondary harm. This “deliver and then exit” strategy is reminiscent of developers intentionally not adding “pay-to-win leaderboards,” using reduced stimulation to suppress potential negative feedback. The problem is that once such services scale up, they will inevitably have to face more detailed rules around standardized procedures, access thresholds, and usage time limits—none of which have yet taken shape.

If we shift our view back to the gaming world, we’re already used to handling “farewell” emotions through character progression, commemorative skins, and commemorative achievements. Players know these carry symbolic meaning, and developers also remind them through event copy that “this is commemoration, not replacement.” Using AI to restore the deceased in real life is essentially turning the “memento” function from low-interaction to high-interaction: a one-off commemorative video is upgraded to a highly interactive character available at any time. The experiential gains from this bump in intensity are very direct, and the risks are amplified accordingly.

For ordinary users, if similar services become normalized in the future, I would be more inclined to see them as “high emotional cost” applications rather than ordinary entertainment products. Much like choosing to commit to a long-running online game, the real key is not how realistic the technology is, but whether you clearly understand your own boundaries of need: do you need companionship for a transitional period, or are you trying to use a never-ending “online save file” to replace a real-life farewell? The technology has already proven that “it can be done.” The next question to answer is “to what extent is it appropriate to do it?” That is not something developers’ self-discipline alone can settle; it will likely require society as a whole to participate in setting the rules.