Stockmark Releases 'Outlier Hypothesis Generation (Beta)' Feature for 'Aconnect' AI Agent to Support Manufacturing R&D

Stockmark Inc., which develops proprietary generative AI foundations and provides generative AI services for businesses, announced the launch of the 'Outlier Hypothesis Generation (Beta)' feature for the technical exploration agent of its manufacturing-focused AI agent, 'Aconnect.'

This feature compensates for the 'fixation of perspective due to expertise' that even seasoned engineers often fall into. By intentionally suggesting wild and unexpected hypotheses that are not in the extension of conventional lines, it expands the boundaries of thinking in R&D and dramatically increases the comprehensiveness of deliberations.

Background: The Deeper the Expertise, the Less Visible the 'Outliers'

In manufacturing R&D, engineers have increased the accuracy of solutions by deepening their expertise. However, in exchange for this mastery, 'convergence of ideas due to expertise'—where thinking is confined to past success patterns or the boundaries of existing technologies—has become a challenge.

Many R&D professionals feel limited in situations such as:

- Convergence toward 'model answer-like hypotheses': Rational but lacking novelty, leading only to safe, conventional plans.
- Unclear definition of comprehensiveness: When asked, 'Are there any other possibilities?', it's hard to define the boundaries of what is considered comprehensive, leading to the mass production of similar ideas.
- Existence of psychological hurdles: In meetings where feasibility and cost are prioritized, 'wild hypotheses' that might hold value are often dismissed before reaching the discussion table.

For staff in divisional R&D facing short-term solutions or headquarters R&D dealing with wide-ranging problem solving, breaking through these 'cognitive biases' on their own was extremely difficult.

Overview of the New 'Outlier Hypothesis Generation (Beta)' Feature

'Aconnect's' technical exploration agent is a tool that presents ideas for problem solving from a vast array of papers and news, allowing users to verify comprehensiveness using logic trees. The newly added 'Outlier Hypothesis Generation (Beta)' feature assists in improving the comprehensiveness of deliberations through a specialized algorithm that AI uses to intentionally extract and present 'perspectives that are not in the extension of existing technologies.'

Features

- Breaking away from 'model answer-like hypotheses': It intentionally avoids 'valid but unsurprising solutions' that experts tend to fall into. It prioritizes the presentation of 'unexpected approaches' not found in conventional technical continuities, such as different technical principles, materials in different fields, or non-traditional processes.
- Visualizing the boundaries of thinking and defining comprehensiveness: In response to anxiety about 'whether there is any further scope for consideration,' AI presents the 'outside of thinking (outliers),' clarifying the edges of the consideration range. This allows users not only to confirm if they have missed anything but also provides a strong evidence-based ground to explain to stakeholders that they have 'made decisions after being this comprehensive.'
- Contribution to novel technical exploration and creation of research themes: It contributes not only to 'expanding the range of options' for short-term tasks like immediate troubleshooting or improvement of existing products but also to the formulation of new research themes. By incorporating perspectives outside the expert domain in long-term theme exploration at headquarters R&D, it becomes possible to stably generate triggers for innovative projects.

Changes brought by this feature: From a 'place for explanation' to a 'place for decision-making'

By having AI remove the wall of 'fixation of perspective' that was unavoidable for experts, dramatic changes will emerge in R&D settings:

- Expansion from 'within the range of experience' to 'outside of thinking': By incorporating outlier perspectives presented by AI in addition to hypotheses based on self/team experience, the width and depth of initial exploration expand overwhelmingly.
- Preparation of 'unexpected points of discussion' before meetings: By grasping 'outliers that were considered but not adopted' in advance, one can objectively and evidence-based answer to pointed questions from superiors such as 'Are there any other options?'
- Cultivating a culture where 'wild ideas' can be 'considered safely': Even wild hypotheses that people might find difficult to voice become easier to bring to the discussion table if presented by AI, vitalizing discussions within the team.
- Breaking away from 'individual-dependent creativity': By turning hypothesis generation, which relied on individual experience, into a process, it enables the maintenance of high exploration capabilities across the entire organization, from junior to senior staff.

About Aconnect

'Aconnect' understands your business, finds information on your behalf, delivers insights, and detects risks and opportunities without missing them. It summarizes necessary information from a wide range of sources, including business news, papers, patents, and internal documents, supporting faster and more reliable judgment in development settings.

- Aconnect: https://aconnect.stockmark.co.jp