KVQ Intellect
Q&A

Questions & Answers

What people most often ask about the picture description task and AI-MindSnap.

01

The picture description task

What is a picture description task, and why can it say anything about cognitive health?

The person looks at a picture of an everyday scene, with several things happening at once, and describes in their own words everything they see going on. There are no right or wrong answers and nothing to memorize; the picture stays on screen the whole time.

Describing a busy scene uses several mental abilities at once: noticing what matters, finding the right words, building sentences and keeping track of what has already been said. Research has found that changes in these abilities can show up in speech — more frequent long pauses, shorter and simpler sentences, vaguer words ("that thing over there"), and fewer of the picture's main events conveyed. That is why picture description, used by neuropsychologists and speech-language therapists for decades, is one of the most studied tasks in speech-based cognitive research. Summaries of key studies are on our References page.

What does AI-MindSnap listen for?

Ten factors, in two groups:

  • Content and language — how many of the picture's key pieces of information were conveyed, how varied the vocabulary is, how complex the sentences are, fluency and hesitations (such as "um" and restarts), and how logical and coherent the description is.
  • Voice and timing — speech rate, how often there are long pauses (2 seconds or more), pitch variation, voice quality, and how clearly words are articulated.

Each factor is measured from the recording itself, and the report shows what each rating was based on.

Why does AI-MindSnap use pictures of everyday life in Taiwan?

The best-known picture in research, the "Cookie Theft", shows a mid-20th-century American kitchen. An unfamiliar setting can make people pause or say less for reasons that have nothing to do with cognition. AI-MindSnap uses five scenes of everyday life in Taiwan — a classroom, a school sports day, a Mid-Autumn Festival barbecue, a Lunar New Year's Eve family dinner and a corner of a park — each with many things happening at once, so older adults describe places and occasions they know. Each description is scored against the content of the picture that was actually shown.

02

Using AI-MindSnap

What is AI-MindSnap, and how is it different from the MMSE or MoCA?

AI-MindSnap (愛腦快篩) is KVQ Intellect's speech-based cognitive pre-screening tool. A person describes a picture for one to three minutes; AI-MindSnap analyzes the content, language, timing and voice of the recording and returns a level of cognitive-decline risk, with a report that can be downloaded as a PDF.

Paper tests such as the MMSE and MoCA are given by trained staff, take ten minutes or more, and ask questions with right and wrong answers. A picture description is a few minutes of natural talking, can be done on a phone without a trained administrator present, and can be repeated with a different picture. AI-MindSnap does not replace those tests: it helps decide who should go on to a full cognitive assessment.

How long does it take, and what equipment is needed?

Speaking takes one to three minutes — at least 30 seconds, and the recording stops by itself at three minutes. The result is usually ready two to three minutes after the recording is uploaded.

All that is needed is a phone, tablet or computer with a microphone, a recent version of Chrome, Edge or Safari, and a stable internet connection; nothing needs to be installed. For tests on site we recommend a tablet on a stand, 30–50 cm from the person.

Can the test be taken at home, or at community events?

Yes. AI-MindSnap can be used:

  • On site, on your device — in a clinic or assessment room.
  • At home, on the person's own phone — staff send a test link or show a QR code. A link is valid for 7 days and can be used for one test.
  • At community events — visitors scan a QR code, register with a code of their own and take the test on their own phones.
  • By uploading a recording made with another device.

At home, the person should sit somewhere quiet, take the test alone and leave the phone on a table. Wherever the test is taken, only the person being tested should speak and nobody should give hints: AI-MindSnap sets other voices aside, but prompting still changes what the person says. Tests given in person by trained staff are the most consistent when results will be compared over time.

Which languages are supported?

Speech analysis is available in 30 language varieties, chosen for the size of their older-adult populations. They include Taiwanese Mandarin, Taiwanese (台語) and Hakka (客語), as well as Mainland Mandarin, Cantonese, English, Japanese, Korean, Vietnamese, Thai, Malay, Indonesian, and major European and Indian languages.

A description in Taiwanese or Hakka is written down word for word in standard Chinese characters, beside a romanization of what was said, and any words the person switches into Mandarin are kept — as many older speakers do. The test page and the report are available in Traditional Chinese, Simplified Chinese and English. Each organization sets the language its tests are analyzed in.

03

Understanding the result

What does a result include?

  • Level of cognitive-decline risk — low, medium or high, following a risk index from 0 to 100: below 40 is low, 40 to 70 is medium, above 70 is high.
  • Estimated MMSE, MoCA and CASI scores — inferred from speech and shown on familiar scales. They are not a substitute for giving those tests.
  • The transcript, the parts of the picture the person mentioned, and the recording for playback. The transcript is produced automatically, so it may contain recognition errors.
  • A full report, online or as a PDF, with your organization's logo.
  • Change over time — each new result is shown next to the previous one and charted across tests. Use a different picture each time, since describing the same picture again can be influenced by memory.

A medium or high result means we recommend a full cognitive assessment, for example by referral to neurology, geriatrics, psychiatry or a memory clinic. A low result does not mean nothing is wrong: anyone still worried about memory or daily functioning should see a doctor.

Is AI-MindSnap a medical device? Can it make a diagnosis?

No to both. AI-MindSnap is not a medical device and has no diagnostic capability; it is meant only as a pre-screening tool. A risk level says how much attention is warranted, not that the person has any condition.

Every result and report carries the statement: "For reference only. This estimate is produced by AI and is not a clinical diagnosis of any kind." Only a qualified clinician, with a full assessment, can make a diagnosis; AI-MindSnap helps decide who may benefit from one.

How accurate is AI-MindSnap?

In our research with Mandarin-speaking participants, our speech analysis has reached accuracy as high as 90%; for comparison, a 2025 review of 54 published studies reported a pooled accuracy of about 80% for speech-based screening.

Results also depend on things other than cognition: slow speech and long pauses can come from tiredness, low mood, hearing difficulty, illness or medication; how much a person says is related to their years of education; and background noise or interruptions affect the recording. The analysis is deliberately cautious, so some people with a medium or high result will turn out to be fine — one reason those results lead to a full assessment, which also looks for treatable causes such as medication side effects, depression, sleep problems or hearing loss.

04

Privacy and data

How is personal data protected?

  • Minimal personal information. Each person is identified by an ID your organization assigns (for example P-1001), with optional age, sex and years of education. Please do not use names, national ID numbers, phone numbers or medical record numbers as IDs.
  • Stored in Taiwan. Records, results and recordings are stored on cloud servers in Taiwan. Parts of the speech analysis are carried out by cloud AI services outside Taiwan, which process the recording to produce the analysis; AI-MindSnap stores nothing there.
  • Seen only by your organization, according to each account's role, with an access log of who viewed or changed which record and automatic sign-out after 20 minutes without activity.
  • Deleted when you choose. Recordings are kept with their results until your organization deletes them. Deleting a test taker removes all of their recordings, transcripts, results and reports at once, and from the last backup copy within eight weeks.
05

For organizations

Who is AI-MindSnap for? Can families use it?

AI-MindSnap is for organizations that work with older adults: hospitals and clinics (including neurology, geriatrics, family medicine and health check-up centers), long-term care and community care centers, community health programs, and partners who want to offer pre-screening in their own app or service.

Individuals and families can try AI-MindSnap through our LINE Official Account, or take it at an organization that offers it — contact us to find one. If you are worried about a family member's memory, please also talk to a doctor.

How is AI-MindSnap priced, and what support is included?

By assessment. Each completed assessment uses one unit of your organization's quota; a recording that cannot be analyzed, or an analysis that fails, is not counted. Administrators can share the quota across the team or set a number for each account, and plans include a set number of user accounts.

Every organization receives a user manual, a step-by-step guide for staff who sit with test takers, a printable instruction card and checklist, and email support. Contact us for pricing.

Can AI-MindSnap connect to our own system?

Yes. Partners can use the integration API to register test takers, create test links, receive results automatically when an analysis finishes, and retrieve reports. Connecting to a hospital information system is scoped project by project; contact us to discuss.

→ Let's talk

Still have a question?

Talk to us about bringing AI-MindSnap to your clinic, care center or community program.