Northeastern researchers propose AI game companions for emotional support
The proposal envisions characters that notice frustration and encourage breaks. Critics question studios’ expertise and incentives—not just the technology.
Loading page…
The proposal envisions characters that notice frustration and encourage breaks. Critics question studios’ expertise and incentives—not just the technology.
Listen to this story
A peer-reviewed paper from Northeastern University London argues that persistent, conversational game characters could be designed to respond to players’ emotional states, not just assist gameplay; the university highlighted the work on Nov. 17. The proposal builds on generative companions already seen in PUBG: Battlegrounds and Where Winds Meet, but emotional monitoring is a possible future use, not a confirmed feature. The debate is whether studios can provide support safely, given concerns about expertise, privacy, harmful advice and emotional dependence.
Melika Vafafar suggests companions could respond to voice cues, while controllers or other sensors might supply physiological signals; these are proposals, not announced game features.
Vafafar says players may spend tens or hundreds of hours gaming in a month, creating repeated opportunities for companions to influence emotions and habits.
Real Safety AI Foundation executive director Travis Gilly argues studios lack the expertise and incentives to test emotional support safely, and questions whether attachment is the real product goal.
Northeastern University London researchers are proposing a new job for video-game AI companions: supporting players’ emotional wellbeing. A peer-reviewed paper, described by the university on November 17, argues that characters able to converse and remember players could offer more than gameplay help. Safety advocates question whether game studios should take on that responsibility.
The paper is titled “Serving the Player or Shaping Them? Game-Embedded AI Companions and the Rewiring of Well-being.” Corresponding author Melika Vafafar, a computer science PhD researcher, argues that increasingly adaptive, emotionally responsive characters could influence how players feel and behave over time—not simply make a game more engaging.
There is already a starting point inside games. Vafafar cited PUBG: Battlegrounds and Where Winds Meet as examples using generative AI for companions or characters. Such companions can appear as avatars, offer tips and travel alongside a player. Her proposal would give those interactions a different purpose: responding to the person’s emotional experience, rather than only helping them progress.
Unlike characters limited to scripted dialogue, generative companions use language models to compose responses in real time. Some can sustain open-ended conversations and remember a player across interactions, Vafafar said. In the proposed support role, a companion might notice frustration in someone’s voice, check in after a losing streak or suggest a break during a long late-night session.
Vafafar also described a more sensor-heavy possibility. Equipment could supply signals beyond what a player says: a controller might detect sweaty palms, while eye tracking, skin-conductance measurements or brain signals could offer physiological cues. These are proposed ways for a companion to interpret a player’s experience, not confirmed features of the games she named.
The argument depends partly on sustained contact. Players may spend tens or hundreds of hours gaming in a month, Vafafar noted, creating repeated opportunities for a companion to affect emotions and habits. She suggested those interactions could have positive effects as a system becomes better at anticipating an individual player’s needs.
They are not yet designed for well-being, but our paper is a provocation about what they could become.
Melika Vafafar, speaking to Northeastern Global News
Travis Gilly, executive director of the Real Safety AI Foundation, challenged the idea that game developers are equipped for that task. He told Northeastern Global News that studios making entertainment lack the relevant expertise and would not pay for the research and testing needed to provide emotional support safely. Gilly also argued that games already offer emotional relief through escapism.
His concern extends to intent. Gilly drew a distinction between building something to support the player and adding a feature because people will become attached to it. That makes the developer’s purpose part of his objection, rather than treating conversational ability alone as sufficient grounds for an emotional-support role.
Amy Bussey, head of insight at the youth gaming-harm prevention charity Ygam, said entertainment, social interaction and emotional support are becoming harder to separate. Her safeguarding concerns include harmful or inaccurate advice, privacy, checks on users’ ages and emotional dependence. She also warned about how AI companions could affect young people’s understanding of relationships and support.
Those concerns sit alongside mixed findings about emotional-support chatbots. Northeastern’s article cites a separate university study reporting potential benefits, including short-term loneliness reduction and improved mood, as well as adverse outcomes such as emotional dependence, worsening symptoms and self-harm. That research concerns chatbots; it is not a demonstration of benefits from the proposed in-game companions.
Bussey nevertheless sees a possible role for AI-powered characters alongside existing support systems, particularly when young people face barriers to getting help elsewhere. Her endorsement is conditional: that opportunity must be balanced against safeguarding risks.
Loading discussion...
Join the conversation
Explain what would make that role appropriate—or inappropriate—for a game developer.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.