Deep Well LabsRESEARCH / INSTRUMENT
Current week
A weekly ledger of feeling, use, relation & meaning.
The public is using AI, marking down the institutions that make it, and beginning to write doctrine about dignity. TEMPER keeps those facts from collapsing into one emoji, with every number’s parent visible.
In the United States, concern about AI in daily life still outruns excitement: 52% more concerned, 37% equally both, 9% more excited. That is a feeling measure, not a use measure, and it is not a score of the world.
Pew: “Since 2024, the share of Americans ages 18 to 34 who are more concerned than excited about AI has jumped from 40% to 55%.” On being equally concerned and excited, Pew writes: “In more than half of the countries surveyed, this is the majority or plurality view among adults under 35.” Youth is not a synonym for excitement.
The partisan sign flipped. Democrats / leaners are now more concerned than Republicans / leaners, 56% to 49% (Pew American Trends Panel, 22–28 Jun 2026). In 2021 the gap ran the other way (31% Dem, 45% Rep). A party cut is not an average of the country.
Pew, across 36 countries with the United States among them: “A median of 40% in high-income countries say they are more concerned than excited about the technology.” In the 18 middle-income countries the median is 31%. Excitement stays low (medians 14% and 13%). A U.S. 52% is not the world, and the world is not English Twitter.
Pew Research Center, American Trends Panel. Field dates 22–28 Jun 2026. N = 3,488 U.S. adults. More concerned in prior years, kept: 50% (2025), 51% (2024), 37% (2021). Published 16 Sep 2026.
People and firms are different populations. Used ChatGPT is not the firm used AI in a business function. Sources sit on the Use view.
Elon University Imagining the Digital Future Center (YouGov field, 18–22 May 2026) measures a use. Pew (25 Aug 2026) measures a public verdict. They are allowed to disagree.
“[T]echnology is never neutral, because it takes on the characteristics of those who devise, finance, regulate and use it.”Leo XIV, Magnifica Humanitas, 15 May 2026. Quoted, not paraphrased. This is a tape of what institutions publish. It is not a percentage of religious people who like AI.
Census BTOS is the ground series. Open Firms by size to swap the source. A McKinsey percentage must not sit in this slot as if it were the Census.
Instrument 1 is Pew’s concerned / excited item only. No composite overall sentiment. Waves are never overwritten.
| Cut | More concerned | Field | Parent |
|---|---|---|---|
| U.S. adults | 52% | 22–28 Jun 2026 | Pew American Trends Panel, N=3,488 |
| Dem / lean Dem | 56% | same wave | Pew short read, 16 Sep 2026 |
| Rep / lean Rep | 49% | same wave | same |
| U.S. ages 18–34 | 55% | same family | Pew Global, 17 Sep 2026 (U.S. young-adult cut) |
| High-income country median | 40% | 8 Feb–13 May 2026 | Pew Global, 17 Sep 2026. 18 high-income countries, U.S. among them. Not a U.S. substitute. |
| Population | Measure | Value | Parent |
|---|---|---|---|
| U.S. adults | Ever use a chatbot | 49% yes / 51% never | Pew ATP, 17–23 Feb 2026, N=5,119 |
| U.S. adults | Daily chatbot use | 24% | same survey |
| U.S. adults, employed | Chatbots for tasks at work | 38% | same survey |
| U.S. firms | Used AI in a business function in prior two weeks | 19.8% | Census BTOS window ending 3 May 2026 |
| U.S. firms, 250+ | same question | 37% | same BTOS window |
Elon University: “27% of U.S. adults who use the internet have significant emotional or social interactions with artificial intelligence large language models (LLMs).” Imagining the Digital Future Center, fielded by YouGov, 18–22 May 2026. Published 2 Sep 2026.
Pew, 25 Aug 2026. The remainder: 14% neither, 27% not sure. A verdict is not a prevalence.
“In practice, however, technology is never neutral, because it takes on the characteristics of those who devise, finance, regulate and use it.”Leo XIV, Magnifica Humanitas, 15 May 2026. English text published by the Holy See.
The sentence above is copied from the source. No paraphrase of the encyclical, by a model or by this page, is printed here. Read the source.
Other traditions are not yet on the tape. An entry is added only when a published primary text can be quoted and linked.
Census Bureau story, 26 May 2026, from BTOS. Question family: whether the business used AI in a business function in the prior two weeks. Census: “AI use increased among firms with at least 20 employees but didn’t change significantly among firms with fewer than 20 employees.” Overall use “hovered between 17% and 20%” from December 2025 to May 2026. This is the default series.
McKinsey-style surveys of large organizations answer a different question about a different frame. TEMPER will cite a specific wave here only with its own N, field dates, and universe. Until that citation is locked, the chip exists to keep the two series from merging.
Optimism on an earnings call is Layer-labeled speech. It is not public opinion and it is not BTOS. Aspect scoring (target, stance, affect, speaker class) belongs to the nowcast when that week is published.
| Layer | What it is | This week |
|---|---|---|
| A Ground truth | Pew, Gallup, Anthropic Public Record, Elon companion survey, Census BTOS. Field dates, question text, N, and margin of error where the source states one. Prior waves kept. | Standing. Figures on this page. |
| B Nowcast | Fixed news panel plus social sample, aspect-scored. Not public opinion. | Panel specified. First scored week not published. A nowcast will never overwrite Layer A. |
| C Human coding | 200–500 items/week, dual-labeled. Automated series need a reliability badge. | Panel is not standing. No badge. No automated headline. |
Checked against these pages on 2026-09-19.
| Week | Status | Note |
|---|---|---|
| Week of 14–18 Sep 2026 | Frozen | Method locked. Layer C not standing. Nowcast unpublished. |
| Earlier weeks | Empty | No back-filled history will be invented. |