Mandy Sha

Research Scientist PMP® Linguist
on LinkedIn 👉

[Mandy Sha @MMandySha]

Book launch social starting 6:00PM May 13, 2026
at the Los Angeles @thebonaventure

🆕📖 𝘊𝘳𝘰𝘴𝘴-𝘊𝘶𝘭𝘵𝘶𝘳𝘢𝘭 𝘙𝘦𝘴𝘦𝘢𝘳𝘤𝘩 𝘪𝘯 𝘗𝘳𝘢𝘤𝘵𝘪𝘤𝘦 of 1⃣9⃣ chapters for improving research quality across cultures, languages, countries 👇🏻
https://t.co/bHYxi2Exmn

[Mandy Sha @MMandySha]

Black turnout doesn't fully explain the '26 midterm election results, how so?

@AndraGillespie @JWJIEmory
"Whites were more likely to vote Democratic in 2026 compared to previous election cycles"

Webinar host: Black #AAPOR group

[Mandy Sha @MMandySha]

What're obs re Black voters '26 midterm election?

@AndraGillespie @JWJIEmory
1⃣ Black voters not too different from voters in general "dissatisfied w the country's direction"
2⃣ Blacks more likely than gen pop to cite the economy = top issue

Webinar host: Black #AAPOR group

[Mandy Sha @MMandySha]

Automated #AI detection for data quality.. what about PEOPLE?

JGoodbread @AskYTM
🏗️ acceptable standard
🧐 edge case, false positive🔻overcleaning
🤔 removed + defend it
👀 quality rpt b4 results

💖 Automation 🔎patterns, openends for AI text signatures + flag score route cases

[Mandy Sha @MMandySha]

How is #AI #DataQuality a 𝘣𝘶𝘴𝘪𝘯𝘦𝘴𝘴 problem (not just technical) in survey research?

J.Goodbread @AskYTM
1⃣ wrong decision
2⃣ reruns, rework = 💸⌛️
3⃣ stakeholder trust expensive to earn, cheap to lose
4⃣ decision velocity: when teams don't trust data, they ask for more

[Mandy Sha @MMandySha]

Survey cost is missing from the literature... how?

@unlbosr K.Olson
❗️ Many costs not documented/shared, eg admin/analysis costs
🧐 Monetary 🆚 nonmonetary [hours] costs
Need 🆕 cost categories beyond fixed, variable, infrastructure costs

👁‍🗨@ESS_Survey : https://t.co/AFx0VFAoK8

[Mandy Sha @MMandySha]

What're examples of 'survey research in flux'?

@unlbosr K.Olson:
⏩changing: modes, methods of inferences, data sources (🔼nonprobability, paradata) but respondents don't always 🫶 them
🤔 Lower cost options may 🔽costs per complete but not overall budget

🙏@ESS_Survey Webinar

[Mandy Sha @MMandySha]

Do #AI agents vary across online survey platforms?

S.Chen @LBS Webinar @SawtoothSoft
yes, some seem to provide data w lower agent rate, 🗝️evasion=knowledge of tests, so:
✔️ layer tests, make it hard to anticipate
📝 edit tests, not all the same
🏗️ detections but NO silver bullet

[Mandy Sha @MMandySha]

What're false positives in #AI detection tests for online surveys?

S.Chen @LBS Webinar @SawtoothSoft:
Humans can fail, too! AI capabilities🆙 but not true of humans❗️

AI detection tests validation for 𝘦𝘧𝘧𝘦𝘤𝘵𝘪𝘷𝘦𝘯𝘦𝘴𝘴: agent consistently fail but humans pass, results:

[Mandy Sha @MMandySha]

What're #AI detection tests that can be executed by indv researchers for online surveys?

S.Chen @LBS Webinar @SawtoothSoft
5 types:
🔍 lack of attention/comprehension
🗣️ admission of AI use
🔍 superhuman abilities
🔍 dif processing abilities
🔍 dif interactions w user interface

[Mandy Sha @MMandySha]

How to control the narrative and answer the question when 🔉insights?

@keenasmustard @ESOMAR
"contextualize", for example:

Add a benchmark to 🗣️ "we've exceeded our customer satisfaction targets", indicate target=30% in the data chart showing about 1/3 customers satisfied

[Mandy Sha @MMandySha]

Why create 🪝 or 1-liner in titles to 🔉insights?

@keenasmustard @ESOMAR
"persuading ppl to 👀 your story, no hook no point"

USE: Surprise "pictures need no passport!"
➕ Curiosity, Question, Negativity, Numbers

🚫 "Results of tracker Q4" "H2 feedback on CX study (attached)"

[Mandy Sha @MMandySha]

What're common challenges product testing w kids?

@curioninsights @MRII_Learn
1⃣ comfort/engagement: reassurance w new exp, limited attention span
2⃣ understand q + describe sensory exp + adult influence
3⃣ update database as kids age out, age-specific logistics: highchair, bib

[Mandy Sha @MMandySha]

What to know about product testing with children?

@curioninsights @MRII_Learn
🧐 body language, emotional cues in qualitative research
🧒kids more definitive, eg. good 🆚 bad
‼️ sensory fatigue
👎 parents influence: Be clear who's 🗣️
👑 adapt methododology to each age group

[Mandy Sha @MMandySha]

The human side of statistical leadership... What to 🤔 in the #AI age?

@AmstatNews panel
1⃣ "Super Intelligence" for fed statistical agencies per 9/29/'26 EO
2⃣ Training entry level statisticians to 🧐 AI on tasks they've never done ⚖️validation 🆚 oversight
⬇️cognitive skills?

[Mandy Sha @MMandySha]

#privacy 🆚 safety, what's the tension 🔎 #AIOLIA_EU @DigitalEU?

@CEA_Officiel A.Grinbaum: "strong anonymization can prevent detection of abusive uses"
limited transparency: overexplaining surveillance ⬇️user trust 🆙attacks
Perceived intrusive? Human oversight > #LLM

🙏@WAPOR

[Mandy Sha @MMandySha]

#AI chatbots & social media now used 🔎election info, what about TRUST?

M.Bitzer @catawbacollege using @YouGov poll:
67% Am. have little or no trust in them🟰voters encounter political info online but don't view the distribution channels as reliable sources

@TFADemocracy cohost

[Mandy Sha @MMandySha]

How confident are North Carolinians about 🗳️elections?

M.Bitzer @catawbacollege using @YouGov poll
"Confidence is strongest where voters have the most direct experience, and it declines as elections move farther from home"

event cohost @TFADemocracy

[Mandy Sha @MMandySha]

How to shift #AI effort to 'high-value work' in stats production?

L.McKeown @ONS webcast @theNASEM
1⃣ classification: generate codes from text
2⃣ survey assist: tailor followup🗣️
3⃣ receipt processing: extract/classify data from scan

👍 Humans to validate + judge reliability!

[Mandy Sha @MMandySha]

What're top 5⃣ training needs according to @OECD_Stat ('20-'25)?

S.MacFeely @OECD webcast @theNASEM
❗️data visualization
❗️web scraping
#AI
❗️data mining
❗️statistical analyses (using @_R_Foundation, @ThePSF)

🗝️= Professional development for official statistics workforce

[Mandy Sha @MMandySha]

What could make #LLM more equitable?

@fraukolos @theNASEM webcast
"fine-tuning on (and human feedback from) representative longitudinal data", needing:
✔️ partnership w academic teams
✔️ continued education
✔️ get benchmark ready from #AI testing