示範交付:多工講解稿來源核對
此為示範樣本。
一個科普 YouTube 頻道請 AI 寫了一篇旁白初稿,共 850 字,交來做來源核對。以下依次是交來的稿、事前確認的範圍、主張來源表、註解,以及未能確認之事項。
委託稿為英文,交付亦以英文寫成。
委託人交來的稿
方括號內的主張編號為核對時所加,原稿沒有。
Why Multitasking Is Making You Worse at Everything
You're answering an email while a video call plays in the background. A chat notification pops up, and you glance at it, nod at something someone said, and type a reply. By the end of the day you feel like you've done a dozen things at once. Here's the uncomfortable part: you probably haven't done any of them well. And the research suggests the people who feel best at multitasking are often the worst at it. [C1]
Start with a word that's quietly misleading. What we call multitasking is, for most tasks, rapid switching. Your brain isn't running two demanding thoughts in parallel; it's flipping between them. Psychologists have studied this since the 1990s. [C2] In a well-known 2001 paper in the Journal of Experimental Psychology: Human Perception and Performance, Joshua Rubinstein, David Meyer, and Jeffrey Evans showed that when people switch between tasks, they lose time every single time they do it. The cost per switch is small, a fraction of a second, but it grows with the complexity of the task, and it adds up fast. [C3]
The same research group framed this as a two-step process: first you shift your goal, then you load the rules for the new task. [C4] That's a real mental operation, and it takes real time. Earlier, in 1995, Harold Pashler at the University of California, San Diego, had described a "bottleneck" in the brain: when two tasks both need a decision, the second one simply waits for the first to finish. [C5]
Now, you might say, "But I'm good at this. I do it all the time." That's exactly what a team at Stanford looked at. In 2009, Eyal Ophir, Clifford Nass, and Anthony Wagner published a study in the Proceedings of the National Academy of Sciences. They compared heavy media multitaskers with light ones. The expectation was that the heavy group would be better at filtering, switching, and juggling. It was the opposite. Heavy multitaskers were more easily distracted by irrelevant information and were slower when switching between tasks. [C6] The people practicing the most were not improving. [C7]
That finding has been debated since. Later studies haven't always reproduced the exact pattern, and researchers disagree about whether heavy multitasking causes attention problems or whether people with weaker attention are simply drawn to multitask. A careful reading says the link is real in many datasets, but the direction of cause is still an open question. [C8]
What about the cost in plain minutes? Gloria Mark, a professor of informatics at the University of California, Irvine, has spent two decades measuring how people work with a screen. In her research, she found that after an interruption it can take around 23 minutes to return to the original task. That figure has been widely quoted, and it comes from observational work on office workers, so treat it as an average of messy real-life situations rather than a law of nature. [C9] Her group also reported that attention on a single screen has shrunk over time, from about two and a half minutes in 2004 to around 47 seconds in more recent measurements. [C10]
Then there's the matter of what interruptions do to your mood. In a 2008 study presented at the ACM Conference on Human Factors in Computing Systems, Mark, Daniela Gudith, and Ulrich Klocke found that people who were interrupted completed their work faster, but paid for it: they reported more stress, frustration, and time pressure. [C11] You can speed up by cutting corners, but your body keeps the receipt. [C12]
Even a phone that isn't doing anything can matter. In 2017, Adrian Ward and colleagues at the University of Texas at Austin published a study in the Journal of the Association for Consumer Research. They found that simply having your smartphone within reach, even face down and silent, reduced available cognitive capacity. The effect was strongest among people who said they were most dependent on their phones. [C13] Some later replication attempts have found weaker results, so this one is best described as promising rather than settled. [C14]
There's also the road. Research from David Strayer and Jason Watson at the University of Utah found that for a small fraction of people, which they nicknamed "supertaskers," dual-task performance doesn't suffer. In their 2010 paper in Psychonomic Bulletin and Review, they estimated that only about 2.5 percent of people fell into that group. For everyone else, talking on the phone while driving made performance worse. [C15] So if you believe you're the exception, the odds say you're probably not.
So what actually helps? The evidence points to boring, practical habits. Work in focused blocks and batch your messages. Turn off notifications that don't need an immediate answer. Put the phone in another room during deep work. [C16] And when you do get interrupted, jot down a one-line note about where you were, which research on resumption suggests can make it easier to pick up the thread. [C17]
The good news is that your brain isn't broken. It's just doing what it was built to do: one demanding thing at a time. Give it that, and you'll likely finish sooner, make fewer mistakes, and end the day less frazzled. [C18]
Sources
- Rubinstein, J. S., Meyer, D. E., & Evans, J. E. (2001). Executive control of cognitive processes in task switching. Journal of Experimental Psychology: Human Perception and Performance.
- Pashler, H. (1994). Dual-task interference in simple tasks: Data and theory. Psychological Bulletin.
- Ophir, E., Nass, C., & Wagner, A. D. (2009). Cognitive control in media multitaskers. Proceedings of the National Academy of Sciences.
- Mark, G., Gudith, D., & Klocke, U. (2008). The cost of interrupted work: More speed and stress. Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI '08), ACM.
- Mark, G. (2023). Attention Span: A Groundbreaking Way to Restore Balance, Happiness and Productivity. Hanover Square Press.
- Ward, A. F., Duke, K., Gneezy, A., & Bos, M. W. (2017). Brain drain: The mere presence of one's own smartphone reduces available cognitive capacity. Journal of the Association for Consumer Research.
- Watson, J. M., & Strayer, D. L. (2010). Supertaskers: Profiles in extraordinary multitasking ability. Psychonomic Bulletin & Review.
事前確認的範圍
| 問題一句 | 這篇旁白稿內的事實主張,各自有沒有來源支持? |
| 邊界 | 核:稿內具名研究、數字、年份、機構與因果說法。不核:寫作風格、影片結構;不改寫全文,不為結論背書。交付只給委託人。 |
| 產出 | 主張來源表、註解稿、未確認事項清單 |
| 備料 | 草稿一份,850 字,四頁,在標準單十頁之內;稿末列出七項來源 |
| 形態與時窗 | 文字往來;確認後五個工作日內交付;交付後可提出一輪書面修訂 |
交付一:主張來源表
18 claims: 6 supported · 9 partly supported · 3 not supported · 0 source not found.
| # | Claim | Source checked | Status | Note |
|---|---|---|---|---|
| C1 | People who feel best at multitasking are often the worst at it | Sanbonmatsu, Strayer, Medeiros-Ward & Watson (2013), PLoS ONE 8(1), e54402 · abstract | Partly supported | Not cited in the draft. The study reports inflated self-rated ability, and that people who multitask most tend to have lower measured ability. It does not report that the highest self-raters score lowest. |
| C2 | Studied since the 1990s | PsyToolKit task-switching page (secondary), citing Jersild (1927), Archives of Psychology 89 | Not supported | Task-switching experiments date to Jersild (1927); the method became widely used from the mid-1990s. Jersild's monograph not read (U4). |
| C3 | Time lost at every switch; small per switch; grows with complexity | Rubinstein, Meyer & Evans (2001), JEP: HPP 27(4), 763–797 · abstract; APA news release, 5 Aug 2001 | Supported | The release gives several tenths of a second for the rule-activation stage. |
| C4 | Two stages: goal shifting, then rule activation | Same | Supported | |
| C5 | Pashler, UC San Diego, 1995, decision bottleneck | Pashler (1994), Psychological Bulletin 116(2), 220–244 · abstract, affiliation | Partly supported | Year is 1994; the draft's own source list says 1994. Affiliation and bottleneck account match. |
| C6 | Ophir, Nass & Wagner (2009): heavy multitaskers more distractible, slower at switching, contrary to expectation | Ophir et al. (2009), PNAS 106(37), 15583–15587 · abstract | Partly supported | Distractibility and worse task-switching performance match; the abstract calls the switching result surprising. "Slower" is not in the abstract (U1). |
| C7 | The people practicing the most were not improving | Same | Not supported | One-time comparison of two groups sorted by a questionnaire index. Neither practice nor improvement was measured. |
| C8 | Replications mixed; causal direction open; link "real in many datasets" | Wiradhany & Nieuwenstein (2017), Attention, Perception, & Psychophysics 79(8), 2620–2641 · abstract; Uncapher & Wagner (2018), PNAS 115(40), 9889–9896 · abstract | Partly supported | Not cited in the draft. Mixed replication and open causal direction match. "Real in many datasets" overstates: the meta-analytic association was weak and lost significance after correction for small-study effects. |
| C9 | Gloria Mark, professor at UC Irvine; about 23 minutes to return to an interrupted task; observational work | Gallup Business Journal interview with Mark, 8 June 2006; Mark, Gudith & Klocke (2008) full text; UC Newsroom, 2 Jan 2024 | Supported (source needs correcting) | The figure (23 min 15 s) is from the 2006 interview: 36 information workers shadowed for three days; average for interrupted work resumed the same day, usually after two other tasks. It is not in the 2008 paper, the only Mark paper in the source list. Mark is now professor emerita. |
| C10 | Attention on one screen: about 2.5 min in 2004, about 47 s recently | UC Newsroom interview with Mark, 2 Jan 2024 | Supported | 47 s averages studies from 2016 to 2020, including replications by other researchers; "her group" is narrower than the source. Book not read (U3). |
| C11 | Mark, Gudith & Klocke (2008): interrupted people faster, more stress, frustration, time pressure | Mark et al. (2008), CHI '08, 107–110 · full text | Supported | Effort was also higher. Lab experiment. |
| C12 | You can speed up by cutting corners | Same | Not supported | The study found no difference in quality between conditions. The sentence contradicts the study it follows. |
| C13 | Ward et al. (2017): phone on the desk, face down, silenced, reduced capacity; strongest for the most dependent; Ward at UT Austin | UT Austin news release, 23 June 2017 (ScienceDaily); record JACR 2(2), 140–154 | Supported (release level) | Desk did worst, another room best; dependence effect held for desk and pocket or bag. Paper not opened (U2); co-author affiliations unchecked (U7). |
| C14 | Later replications found weaker results | Ruiz Pardo & Minda (2022), Acta Psychologica 230, 103717 · abstract | Partly supported | Understated: this preregistered direct replication of Ward et al.'s second experiment found no effect of phone location on either measure. |
| C15 | Watson & Strayer (2010), Utah: about 2.5% supertaskers; phone talk while driving worsened performance for others | Watson & Strayer (2010), Psychonomic Bulletin & Review 17(4), 479–485 · abstract | Partly supported | 2.5% of 200 participants, one sample. The concurrent task was an auditory working-memory task (OSPAN) while driving in a simulator; delivery by phone unconfirmed (U8). The record lists Watson at the University of Utah; Strayer's affiliation is not in the record (U7). |
| C16 | Evidence points to focus blocks, batching, notifications off, phone in another room | Kushlev & Dunn (2015), Computers in Human Behavior 43, 220–228 · title only; C13, C14 | Partly supported | No source in the draft. The title reports less frequent email checking reduces stress (stress, not output); abstract not read (U6). "Another room" rests on C13, which failed one replication (C14). Focus blocks and notifications: no source checked. |
| C17 | A one-line note on where you were helps you pick up the thread | Leroy & Glomb (2018), Organization Science 29(3), 380–397 · abstract | Partly supported | A brief plan for returning reduced attention residue and protected performance on the interrupting task. Faster return to the original task is not reported there; Trafton et al. (2003) record only (U5). |
| C18 | One thing at a time: finish sooner, fewer mistakes | Monsell (2003), Trends in Cognitive Sciences 7(3), 134–140 · abstract; C11 | Partly supported | More errors right after a switch matches. "Finish sooner" conflicts with C11, where interrupted people finished faster. |
交付二:註解稿
註解按主張編號對回上文草稿。
- ¶1 [C1] Add Sanbonmatsu et al. (2013) and narrow the line: people who multitask most tend to be least able to, and most people overrate their own ability.
- ¶2 [C2] "since the 1990s" → "since the 1920s; the modern method took shape in the 1990s".
- ¶3 [C5] 1995 → 1994, to match the source list.
- ¶4 [C6] "slower when switching" → "worse on a task-switching test", unless the full paper is checked. [C7] Cut, or say the study compared two groups once and cannot show whether practice helps.
- ¶5 [C8] Add both sources. "the link is real in many datasets" → "the link is small and has not held up consistently".
- ¶6 [C9] Cite the 2006 Gallup interview, not the 2008 paper. Say people got back to interrupted work about 23 minutes later, often after two other tasks. "professor" → "professor emerita". [C10] "Her group" → "her research and later replications".
- ¶7 [C12] Cut "by cutting corners": the study found no loss in quality.
- ¶8 [C14] "weaker results" → "a 2022 replication found no effect", with the citation.
- ¶9 [C15] "2.5 percent of people" → "2.5 percent of 200 participants". Describe the second task as the paper does after checking it (U8).
- ¶10 [C16] Add sources, or soften "the evidence points to". [C17] Cite Leroy & Glomb (2018) and match its finding: a quick plan for getting back helped people focus on the interruption itself.
- ¶11 [C18] Reconcile with ¶7: interrupted people were faster but more stressed.
- Sources list Add volume, issue and pages (as in §3); add the sources for C1, C8, C9, C14, C17; cite the 2023 book only for points checked in it, with page numbers.
交付三:未確認事項
| # | Item | Why not confirmed |
|---|---|---|
| U1 | Ophir et al. (2009): "slower" at switching | Full text not accessible in this check; abstract says "worse" |
| U2 | Ward et al. (2017) paper text | Publisher page and university repository refused access; release used |
| U3 | Mark (2023), Attention Span | Book not read |
| U4 | Jersild (1927) | Original not read; date from a secondary source |
| U5 | Trafton et al. (2003) | Bibliographic record only |
| U6 | Kushlev & Dunn (2015) | Bibliographic record only |
| U7 | Affiliations of Nass, Wagner, Strayer and Ward's co-authors | Source records list only Ophir (Stanford) and Watson (Utah) |
| U8 | Watson & Strayer (2010): was the auditory task delivered by phone | Abstract silent |
| U9 | Ruiz Pardo & Minda (2022): sample size | Not stated in the abstract |
核對:Ansel · 核對日期:2026-11-08
範圍、備料與價格見〈合作與洽詢〉▸