AI, EQ & Communication News — 2026-10-28 (Wed) Daily News

Overview

Voice and text AI developers pushed emotional intelligence (EQ) as a design target. NTT and NTT Docomo announced EQ-focused dialogue technologies, and Tech Xplore reported a framework for more empathetic LLMs. On the human side, Adecco Group's coaching unit EZRA and Ojin announced coaching and role-play offerings, while former Yahoo CEO Marissa Mayer drew criticism over a personal use of AI.

Emotion-reading AI

- A bumble bee-inspired AI is described as reading students' facial emotions to support English learning (Bioengineer.org). - An AI is described as learning to hear emotion in solo violin through contrastive self-teaching (Bioengineer.org). 【AI empathy and emotional intelligence】 - A new neural framework could make LLMs more empathetic (Tech Xplore). - NTT Docomo developed a framework to improve emotional intelligence in voice AI (Telecompaper). 【Human EQ and strengths in the AI era】 No notable developments. 【Changes in conversation and relationships】 - NTT developed a two-way voice dialogue AI technology that enables natural, human-like conversation (Cloud Watch).

Words written and spoken by AI

- Former Yahoo CEO Marissa Mayer asked AI for 3 words describing what made her daughter special after drawing a blank (The Cool Down). - Mayer faced flak after asking a chatbot to describe her daughter's traits; Livemint's headline asks "'Perfect' answer from AI?" (Livemint). Wonderwall also reported backlash over the AI use (Wonderwall).

Speaking practice and coaching

- Ojin launched Human+, an AI role-play training platform where employees practice hard workplace conversations with lifelike AI characters (Yahoo Finance). - EZRA, the coaching unit of Adecco Group, announced Loop, billed as the first coaching experience combining humans and AI (Kyodo PR Wire).

Research and surveys

No notable developments. 【Japan's EQ, communication and AI】 - NTT developed a technology that improves AI responses from an EQ perspective. The score on a public EQ test rose by about 23% (IT Leaders). - NTT's high-EQ dialogue technology targets the problem of AI sycophancy. Evaluation scores rose by a little over 20% on average (Business Network). - NTT announced a high-EQ dialogue intelligence technology that deeply understands the situation behind words to produce better responses. It aims to deploy the technology in smartphone chat services so that AI can be used with confidence over the long term (Digital PR Platform). - NTT's dialogue AI is reported to suppress excessive agreement. The situation-understanding score improved by 23%, and a demonstration with Docomo is planned (BigGo Finance). - A further outlet described the same NTT work as a major improvement in AI response ability (News Media VOIX).

Statements by key figures

No notable developments. 【Integrated analysis: what is happening】 - The NTT items, the NTT Docomo framework and the Tech Xplore report all treat emotional quality in AI responses as something to engineer and score. NTT's reports name sycophancy and excessive agreement as the targeted weakness, which suggests the goal is shifting from responses that please users to responses that fit the situation. This is an inference from the headlines. - Voice is a second channel for this work: NTT Docomo's voice AI framework and NTT's two-way voice dialogue technology appear to be developed in parallel with text-based efforts. Bioengineer.org's emotion-recognition items (facial expressions, violin) show the reading side advancing in other media. - Coaching and training products (Ojin's Human+, EZRA's Loop) put AI into the practice of human conversation, and Loop is described as combining humans and AI. The Mayer reports suggest public tolerance for AI-generated personal wording may be limited, though this rests on headlines about one case. - For readers weighing these claims: the NTT figures are reported differently (about 23% on a public EQ test, a little over 20% on average on evaluation scores, 23% on a situation-understanding score). The headlines do not show whether these are the same measure, so the figures should not be treated as interchangeable. Improvements on such scores also do not establish that AI understands emotion.

What to watch

- Whether NTT and Docomo's planned demonstration leads to deployment in smartphone chat services, and which evaluation measures are disclosed. - Whether the human-AI coaching and role-play offerings from EZRA and Ojin publish results on how they affect conversation skills. - Whether public criticism like that directed at Mayer shapes how AI-written personal messages are received.