Bengkel Kajian Literatur AIAI Literature Review Workshop
Cari, urus dan tulis tinjauan literatur dengan lebih pantas dan beretika.Search, manage and write literature reviews faster and ethically.
Tiga jam, tiga kemahiran, satu portfolioThree hours, three skills, one portfolio
Memperkasakan kemahiran mencari, menganalisis dan mengaplikasikan literatur AI secara sistematik demi penyelidikan yang lebih berkualiti dan berimpak tinggi.Strengthen skills to search, analyse and apply literature systematically with AI, for higher-quality, higher-impact research.
Anda keluar dengan 5 artifak kajian literatur yang boleh terus digunakanYou leave with 5 literature-review artefacts you can use tomorrow
Soalan + rentetan carianQuestion + search string
Soalan berkerangka PEO/SPIDER dan rentetan Boolean yang diuji di Scopus.A PEO/SPIDER-framed question and a Boolean string tested in Scopus.
Peta literaturLiterature map
Rangkaian sitasi daripada 1–2 kertas benih, dengan senarai rujukan yang disahkan.A citation network from 1–2 seed papers, with a verified reference list.
Log saringan mini-PRISMAMini-PRISMA screening log
Kriteria inklusi/eksklusi dan carta alir PRISMA 2020 berangka.Inclusion/exclusion criteria and a numbered PRISMA 2020 flow.
Matriks ekstrak & temaExtraction matrix & themes
Jadual 5 kajian dengan petikan sumber, dan peta jurang.A 5-study table with source quotes, and a gap map.
Perenggan sintesis + pendedahan AISynthesis paragraph + AI disclosure
Satu perenggan sintesis dan kenyataan penggunaan AI ikut Garis Panduan UKM.One synthesis paragraph and an AI-use statement following the UKM Guideline.
Setiap sesi menambah satu kotak dalam Portfolio Kajian Literatur anda. Kita semak semula senarai ini pada 11.55 pagi.Each session adds a piece to your Literature Review Portfolio. We revisit this checklist at 11.55 am.
Nilai sebagai kompas, rangkaian sebagai petaValues as the compass, networks as the map
Iman · Ilmu · AmalFaith · Knowledge · Practice
Setiap teknik dikaitkan dengan satu nilai: amanah (sitasi tepat), integriti (dedahkan AI), kebertanggungjawaban (manusia membuat keputusan akhir).Every technique is tied to a value: trustworthiness (accurate citations), integrity (disclose AI), accountability (humans make the final call).
Ilmu hidup dalam rangkaianKnowledge lives in networks
Literatur ialah rangkaian sitasi, pengarang dan idea. Kemahiran utama: tahu di mana mencari dan melihat hubungan antara nod.Literature is a network of citations, authors and ideas. The core skill: knowing where to look and seeing connections between nodes.
Lihat → Cuba → Semak → KongsiSee → Try → Check → Share
Demo alat AI, latihan amali di komputer makmal, kuiz dengan maklum balas segera, kemudian kongsi dapatan dengan rakan sebelah.Tool demo, hands-on at the lab PC, a quiz with instant feedback, then share findings with your neighbour.
Siemens, G. (2005). Connectivism: A learning theory for the digital age. Int. J. Instructional Technology & Distance Learning, 2(1). itdl.org/Journal/Jan_05/article01.htm
Bahagian mana kajian literatur yang paling memakan masa anda?Which part of a literature review costs you the most time?
Klik pilihan untuk menambah satu undi. Klik bar untuk tolak satu.Click an option to add one vote. Click a bar to remove one.
Apa pun jawapan anda, prinsipnya sama: AI mempercepat langkah mekanikal; penilaian kekal milik anda.Whatever your answer, the principle is the same: AI speeds up the mechanical steps; judgement stays yours.
AI boleh mempercepat ulasan bertahun-tahun, dan juga mencipta rujukan palsuAI can compress years of review work, and it can also invent references
Sistem otto-SR menghasilkan semula dan mengemas kini 12 ulasan Cochrane dalam 2 hari, anggaran ±12 tahun-kerja manusia. Sensitiviti saringan 96.7% berbanding 81.7% saringan dua manusia.The otto-SR system reproduced and updated 12 Cochrane reviews in 2 days, an estimated ~12 human work-years. Screening sensitivity 96.7% vs 81.7% for dual human screening.
Cao, C., … Bobrovitz, N. (2025). Automation of systematic reviews with large language models. medRxiv (pracetak, belum dinilai rakan / preprint, not peer-reviewed). doi:10.1101/2025.06.13.25329541
Rujukan rekaan dalam tinjauan literatur yang ditulis GPT-3.5 dan GPT-4. Pada 2026, ejen deep research masih menghasilkan 3–13% URL sitasi yang tidak pernah wujud.Fabricated references in literature reviews written by GPT-3.5 and GPT-4. In 2026, deep-research agents still produce 3–13% citation URLs that never existed.
Walters, W. H., & Wilder, E. I. (2023). Scientific Reports, 13, 14045. doi:10.1038/s41598-023-41032-5 · Rao, D., Wong, E., & Callison-Burch, C. (2026). arXiv:2604.03173. arxiv.org/abs/2604.03173
Garis Panduan GenAI UKM (2025): pelajar kekal bertanggungjawab sepenuhnya atas semua kandungan, dan setiap sitasi mesti disemak sendiri dengan sumber asal.UKM GenAI Guideline (2025): the student remains fully accountable for all content, and every citation must be personally checked against the original source.
CARISEARCH
Dari soalan kajian kepada rentetan carian yang diuji, peta sitasi, dan senarai rujukan yang disahkan.From research question to a tested search string, a citation map, and a verified reference list.
🎯 Artifak 01 + 02: soalan + rentetan carian, dan peta literatur.🎯 Artefacts 01 + 02: question + search string, and a literature map.
Jenis kajian menentukan tahap ketelitian, dan alat AI yang sesuaiThe review type sets the rigour, and which AI tools fit
Gambaran luas dan latar konsep, contohnya Bab 2 tesis atau artikel konsep.Broad overview and conceptual background, e.g. a thesis Chapter 2 or a concept paper.
Fleksibel; tiada protokol wajib. Risiko berat sebelah pemilihan tinggi, jadi nyatakan strategi carian.Flexible; no mandatory protocol. High selection-bias risk, so state your search strategy.
Consensus · Elicit · NotebookLM · ResearchRabbit
Memetakan saiz dan jenis bukti dalam bidang baharu; kenal pasti jurang.Map the extent and type of evidence in an emerging field; identify gaps.
Rangka Arksey & O'Malley; laporan PRISMA-ScR. Tiada penilaian kualiti wajib.Arksey & O'Malley framework; PRISMA-ScR reporting. No mandatory quality appraisal.
Scopus AI · Litmaps · ASReview · Elicit
Menjawab soalan khusus dengan semua bukti yang layak; boleh disertai meta-analisis.Answer a focused question with all eligible evidence; may include meta-analysis.
Protokol berdaftar, saringan dua penilai, PRISMA 2020 + PRISMA-S. Tahap tertinggi.Registered protocol, dual screening, PRISMA 2020 + PRISMA-S. Highest rigour.
Rayyan · ASReview · Covidence · Elicit Systematic Review
Menggabungkan kajian empirikal dan teoretikal untuk membina kerangka baharu.Combine empirical and theoretical work to build a new framework.
Sederhana–tinggi; kekuatan pada sintesis kritis (Snyder, 2019).Moderate–high; strength lies in critical synthesis (Snyder, 2019).
NotebookLM · scite · Gemini / Claude (pengkritik)
Menganalisis corak penerbitan: pengarang, negara, rangkaian kata kunci.Analyse publication patterns: authors, countries, keyword networks.
Tinggi pada data; bergantung kepada eksport Scopus/WoS yang bersih.High on data; depends on clean Scopus/WoS exports.
VOSviewer · Bibliometrix · Scopus AI · Gemini (kod R/Python)
Grant, M. J., & Booth, A. (2009). A typology of reviews. Health Info Libr J, 26(2), 91–108. doi:10.1111/j.1471-1842.2009.00848.x · Snyder, H. (2019). Literature review as a research methodology. J Business Research, 104, 333–339. doi:10.1016/j.jbusres.2019.07.039 · Arksey, H., & O'Malley, L. (2005). doi:10.1080/1364557032000119616
Enam peringkat, enam titik semak manusiaSix stages, six human checkpoints
Mempercepat langkah berulang: sinonim, saringan awal, isi jadual, draf struktur.Speeds up repetitive steps: synonyms, first-pass screening, table filling, outline drafts.
Menetapkan soalan, kriteria, tafsiran dan hujah. Keputusan akhir di tangan penyelidik.Set the question, criteria, interpretation and argument. The final decision is the researcher's.
Kenyataan bersama Cochrane, Campbell, JBI & CEE (2025): AI digunakan dengan pengawasan manusia; penyelidik bertanggungjawab sepenuhnya.Joint statement by Cochrane, Campbell, JBI & CEE (2025): AI is used with human oversight; researchers remain ultimately responsible.
Flemyng, E., Noel-Storr, A., Macura, B., et al. (2025). Position statement on AI use in evidence synthesis. JBI Evidence Synthesis, 23(11), 2162–2166. doi:10.11124/JBIES-25-00480
Pilih alat ikut tugas, bukan ikut popularitiChoose tools by task, not popularity
Scopus AI · Elicit · Consensus · Semantic Scholar
Mencari dalam pangkalan kertas sebenar. Jawapan berpaut kepada sumber.Search a corpus of real papers. Answers link to sources.
Risiko halusinasi: rendahHallucination risk: lowResearchRabbit · Litmaps · Connected Papers
Bermula dari kertas benih, menemui kerja berkaitan melalui sitasi. Konektivisme dalam amalan.Start from seed papers and discover related work through citations. Connectivism in practice.
Risiko: rendahRisk: lowNotebookLM · SciSpace · scite
Bertanya kepada PDF anda sendiri. Jawapan berpaut kepada petikan dalam dokumen.Ask questions of your own PDFs. Answers cite passages in the documents.
Risiko: sederhanaRisk: moderateGemini · ChatGPT · Claude · Copilot
Hebat untuk sinonim, struktur, kritikan. Jangan jadikan sumber rujukan.Great for synonyms, structure, critique. Never use as a source of references.
Risiko rujukan: tinggiReference risk: highGunakan akaun UKM: Scopus dan Web of Science melalui laman PTSL, Gemini melalui Google Workspace UKM (data tidak digunakan untuk latihan model, semak terma semasa). Untuk literatur tempatan: MyCite dan MyJurnal.Use your UKM account: Scopus and Web of Science via the PTSL portal, Gemini through UKM Google Workspace (check current terms on data use). For local literature: MyCite and MyJurnal.
Sebelum Scopus, ada laci kad. Ideanya masih sama.Before Scopus there were card drawers. The idea hasn't changed.
George Boole menerbitkan The Laws of Thought: algebra di sebalik AND, OR, NOT.George Boole publishes The Laws of Thought: the algebra behind AND, OR, NOT.
Eugene Garfield mencadangkan indeks sitasi dalam Science: idea asal peta sitasi.Eugene Garfield proposes a citation index in Science: the origin of citation maps. doi:10.1126/science.122.3159.108
Science Citation Index diterbitkan: carian sitasi bercetak.The Science Citation Index is published: citation search in print.
Scopus dan Google Scholar dilancarkan pada bulan yang sama: November 2004.Scopus and Google Scholar launch in the same month: November 2004.
ChatGPT dilancarkan (30 Nov): soalan bahasa biasa menjadi antara muka carian.ChatGPT launches (30 Nov): plain-language questions become a search interface.
Ejen penyelidikan merancang, mencari dan memantau sendiri. Boole masih di dalamnya.Research agents plan, search and monitor on their own. Boole is still inside.
Pecahkan soalan kepada blok konsep sebelum bertanya kepada AIBreak the question into concept blocks before asking AI
Contoh tempatan: "Bagaimanakah tekanan kewangan mempengaruhi kesejahteraan psikologi belia B40 di bandar Malaysia?"Local example: "How does financial stress affect the psychological well-being of urban B40 youth in Malaysia?"
Belia 18–30, isi rumah B40, bandarYouth 18–30, B40 households, urban
youth* · "young adult*" · "emerging adult*" · belia
Tekanan kewangan / kos sara hidupFinancial stress / cost of living
"financial stress" · "economic hardship" · "low income" · B40
Kesejahteraan psikologiPsychological well-being
"psychological well-being" · "mental health" · "life satisfaction" · kesejahteraan
Ibu tunggal B40, Lembah KlangB40 single mothers, Klang Valley
Pengalaman mengurus tekanan kewanganExperience of managing financial stress
interview* · "focus group*" · phenomenolog*
"coping" · "lived experience*" · resilien*
qualitative · "mixed method*"
Pelajar universiti B40B40 university students
Program literasi kewanganFinancial-literacy programme
Tiada program / program biasaNo programme / usual practice
Tekanan psikologi (DASS-21)Psychological distress (DASS-21)
Cooke, A., Smith, D., & Booth, A. (2012). Beyond PICO: The SPIDER tool for qualitative evidence synthesis. Qualitative Health Research, 22(10), 1435–1443. doi:10.1177/1049732312452938
Minta AI cadangkan sinonim; anda yang menguji di ScopusAsk AI for synonyms; you test them in Scopus
Anda pustakawan penyelidikan. Bina rentetan carian Boolean untuk Scopus (TITLE-ABS-KEY) bagi soalan: "Bagaimanakah tekanan kewangan mempengaruhi kesejahteraan psikologi belia B40 di bandar Malaysia?" Guna kerangka PEO. Beri 3–5 sinonim setiap blok, guna tanda petik untuk frasa dan * untuk pemangkasan. Jangan senaraikan sebarang artikel.You are a research librarian. Build a Boolean search string for Scopus (TITLE-ABS-KEY) for the question: "How does financial stress affect the psychological well-being of urban B40 youth in Malaysia?" Use the PEO framework. Give 3–5 synonyms per block, use quotes for phrases and * for truncation. Do not list any articles.
TITLE-ABS-KEY ( ( youth* OR "young adult*" OR "emerging adult*" ) AND ( "financial stress" OR "economic hardship" OR "low income" OR b40 OR povert* ) AND ( "psychological well-being" OR "mental health" OR "life satisfaction" OR depress* ) )AND TITLE-ABS-KEY ( malaysia* )AND PUBYEAR > 2015AND ( LIMIT-TO ( DOCTYPE , "ar" ) )
Terjemahkan rentetan tadi kepada istilah Bahasa Melayu untuk carian di MyCite dan Google Scholar. Kekalkan struktur P AND E AND O. Cadangkan juga istilah yang lazim dalam jurnal tempatan seperti Akademika dan e-Bangi.Translate the previous string into Malay terms for searching MyCite and Google Scholar. Keep the P AND E AND O structure. Also suggest terms common in local journals such as Akademika and e-Bangi.
Kenapa? Literatur tempatan berbahasa Melayu sering tercicir jika hanya carian bahasa Inggeris digunakan.Why? Malay-language local literature is often missed by English-only searches.
( belia OR "golongan muda" OR "dewasa muda" )AND ( "tekanan kewangan" OR "kos sara hidup" OR B40 OR "berpendapatan rendah" )AND ( "kesejahteraan psikologi" OR "kesihatan mental" OR "kepuasan hidup" )
Tukar rentetan Scopus tadi ke sintaks Web of Science (TS=). Gunakan operator proksimiti NEAR/3 untuk "financial" dan "stress". Terangkan setiap perubahan dalam satu baris supaya saya boleh melaporkannya ikut PRISMA-S.Convert the Scopus string to Web of Science syntax (TS=). Use the NEAR/3 proximity operator for "financial" and "stress". Explain each change in one line so I can report it per PRISMA-S.
Proksimiti menangkap "stress caused by financial…" yang terlepas oleh frasa tepat.Proximity catches "stress caused by financial…" that an exact phrase misses.
TS=( youth* OR "young adult*" OR "emerging adult*" )AND TS=( (financ* NEAR/3 (stress* OR strain OR hardship)) OR "low income" OR b40 )AND TS=( "psychological well-being" OR "mental health" OR depress* )AND CU=( Malaysia )# Dicari / searched: 20-10-2026 · n = ___
Kajian kualitatif (SPIDER) menambah blok reka bentuk dan menukar hasil kepada pengalaman. Kajian kuantitatif menambah blok kaedah seperti survey atau regression.A qualitative review (SPIDER) adds a design block and turns outcomes into experiences. A quantitative review adds a methods block such as survey or regression.
Ubah rentetan ini untuk sintesis bukti kualitatif menggunakan SPIDER. Tambah blok reka bentuk (temu bual, kumpulan fokus, fenomenologi) dan blok penilaian (pengalaman hidup, daya tindak).Adapt this string for a qualitative evidence synthesis using SPIDER. Add a design block (interviews, focus groups, phenomenology) and an evaluation block (lived experience, coping).
TITLE-ABS-KEY ( ( "single mother*" OR "lone mother*" OR "female-headed household*" ) AND ( "financial stress" OR "economic hardship" OR povert* ) AND ( interview* OR "focus group*" OR phenomenolog* OR qualitative ) AND ( "lived experience*" OR coping OR resilien* ) )AND TITLE-ABS-KEY ( malaysia* )
Enjin carian AI menjawab dengan sitasi, tetapi anda tetap membacaAI search engines answer with citations, but you still read
LLM terkini mencapai sensitiviti tinggi (0.85–0.94) dalam saringan, tetapi spesifisiti berbeza ikut model [1].Recent LLMs reach high sensitivity (0.85–0.94) in screening, but specificity varies by model [1].
Prestasi bergantung kepada prompt dan bahasa; pengesahan manusia masih disyorkan [2].Performance depends on prompts and language; human validation is still recommended [2].
Saringan pembelajaran aktif mengurangkan beban kerja dengan menyusun rekod paling relevan dahulu [3].Active-learning screening cuts workload by ranking the most relevant records first [3].
⚠ Jurang: sedikit kajian dalam sains sosial bukan bahasa InggerisGap: few studies in non-English social science
| # | KertasPaper | Dapatan utamaKey finding | DOI |
|---|---|---|---|
| 1 | Oami, Okada & Nakada (2025) · Res Synth Methods | 4 LLM dibanding; GPT-4o spesifisiti 0.97, Gemini 1.5 Pro sensitiviti 0.944 LLMs compared; GPT-4o specificity 0.97, Gemini 1.5 Pro sensitivity 0.94 | 10.1017/rsm.2025.10014 |
| 2 | Khraisha et al. (2024) · Res Synth Methods | GPT-4 untuk saringan & ekstrak pelbagai bahasa: hasil bercampur, perlu berhati-hatiGPT-4 for multilingual screening & extraction: mixed results, use with caution | 10.1002/jrsm.1715 |
| 3 | van de Schoot et al. (2021) · Nature Mach Intell | ASReview: alat sumber terbuka pembelajaran aktif untuk saringanASReview: open-source active-learning tool for screening | 10.1038/s42256-020-00287-7 |
Langkah anda: buka setiap DOI, baca abstrak, dan semak sama ada ringkasan AI benar-benar disokong oleh kertas itu.Your move: open each DOI, read the abstract, and check whether the AI summary is actually supported by the paper.
Satu kertas benih membuka seluruh kejiranan ilmuOne seed paper opens a whole knowledge neighbourhood
Page et al. (2021)
The PRISMA 2020 statement. BMJ, 372, n71.
Kertas benih. Garis panduan pelaporan kajian sistematik terkini.Seed paper. The current reporting guideline for systematic reviews.
doi:10.1136/bmj.n71- Saiz nod ≈ pengaruh (sitasi)Node size ≈ influence (citations)
- Warna = kelompok temaColour = topic cluster
- Garis = hubungan sitasi (sahkan dalam alat sebenar)Line = citation link (confirm in the real tool)
"Rujukan Terlalu Sempurna""The Too-Perfect References"
Setiap rujukan lalu 4S sebelum masuk ke ZoteroEvery reference passes 4S before it enters Zotero
DOI menyelesai di doi.org; rekod ada di Scopus, WoS, Google Scholar atau MyCite.DOI resolves at doi.org; the record is in Scopus, WoS, Google Scholar or MyCite.
Pengarang, tahun, jurnal, jilid dan halaman sama dengan rekod Crossref.Authors, year, journal, volume and pages match the Crossref record.
Anda telah membaca bahagian yang menyokong dakwaan. scite menunjukkan konteks sitasi: menyokong atau menyanggah.You have read the passage that supports the claim. scite shows citation context: supporting or contrasting.
Tidak ditarik balik: semak notis retraksi di laman penerbit atau pangkalan Retraction Watch.Not retracted: check retraction notices on the publisher page or the Retraction Watch database.
Rahman, A., & Lim, S. (2022). Financial stress and psychological well-being of B40 youth in the Klang Valley. Malaysian Journal of Social Inquiry, 18(2), 45–62. https://doi.org/10.9999/mjsi.2022.018
Snyder, H. (2018). Literature review as a research methodology: An overview and guidelines. Journal of Business Research, 104, 333–339.
Tiga soalan, satu minit setiap satuThree questions, one minute each
Bina Artifak 01 + 02 untuk topik anda sendiriBuild Artefacts 01 + 02 for your own topic
Tulis soalan kajian anda dan pecahkan kepada PEO atau SPIDER.Write your research question and split it into PEO or SPIDER.
Gemini (akaun UKM): salin prompt Demo 1, tukar soalan, jana rentetan.Gemini (UKM account): copy the Demo 1 prompt, swap in your question, generate the string.
Jalankan di Scopus melalui PTSL. Catat jumlah rekod setiap blok. Laraskan sinonim.Run it in Scopus via PTSL. Record hits per block. Adjust synonyms.
Pilih 1 kertas benih → ResearchRabbit/Litmaps → simpan 5 kertas berkaitan.Pick 1 seed paper → ResearchRabbit/Litmaps → save 5 related papers.
Jalankan 4S pada 3 rujukan. Kongsi satu penemuan dengan rakan sebelah.Run 4S on 3 references. Share one finding with your neighbour.
Tersekat? Angkat tangan, atau tanya pustakawan PTSL di bilik ini.Stuck? Raise a hand, or ask the PTSL librarians in the room.
Carian sedang berubah daripada kata kunci kepada ejenSearch is moving from keywords to agents
Deep Research
Ejen AI membina pelan kajian, mencari dalam Scopus, menyesuaikan strategi, dan menghasilkan laporan dengan skor keyakinan.An AI agent drafts a research plan, searches Scopus, adapts its strategy, and writes a report with confidence scores.
Elicit Research Agent & Routines
Ejen penyelidikan (Ogos 2026) dan Routines (30 Sep 2026) yang memantau bukti baharu mengikut jadual.A research agent (Aug 2026) and Routines (30 Sep 2026) that monitor new evidence on a schedule.
API & MCP
Elicit membuka API dan pelayan MCP: ejen lain boleh memanggil carian akademik secara terus. Aliran kerja kajian menjadi boleh disambung.Elicit opened an API and MCP server: other agents can call academic search directly. Review workflows become composable.
urlhealth
Alat sumber terbuka semak URL sitasi. Apabila model membetulkan diri dengannya, sitasi tidak sah turun ke <1%.Open-source citation-URL checker. When models self-correct with it, non-resolving citations drop below 1%.
Rao et al. (2026). arXiv:2604.03173
Implikasi: lebih banyak automasi bermakna lebih banyak tanggungjawab verifikasi, bukan kurang.Implication: more automation means more verification responsibility, not less.
Minum pagiMorning break
Cabaran rehat: cari satu rujukan dalam nota lama anda yang gagal 4S. Kongsi selepas rehat.Break challenge: find one reference in your old notes that fails 4S. Share after the break.
URUS & ANALISISMANAGE & ANALYSE
Dari ratusan rekod kepada set kajian terpilih, jadual ekstrak berpaut sumber, dan peta jurang yang boleh dipertahankan.From hundreds of records to a selected set, a source-grounded extraction table, and a gap map you can defend.
🎯 Artifak 03 + 04: log saringan mini-PRISMA, matriks ekstrak dan tema.🎯 Artefacts 03 + 04: mini-PRISMA screening log, extraction matrix and themes.
Setiap rekod mesti boleh dijejak: dari mana masuk, kenapa keluarEvery record must be traceable: where it came in, why it left
Alat sebenar: aplikasi Shiny PRISMA2020 (Haddaway et al., 2022) menjana carta yang boleh diterbitkan.Real tool: the PRISMA2020 Shiny app (Haddaway et al., 2022) generates publishable diagrams.
Page, M. J., et al. (2021). The PRISMA 2020 statement. BMJ, 372, n71. doi:10.1136/bmj.n71 · Haddaway, N. R., et al. (2022). PRISMA2020: An R package and Shiny app. Campbell Syst Rev, 18, e1230. doi:10.1002/cl2.1230 · Rethlefsen, M. L., et al. (2021). PRISMA-S. doi:10.1186/s13643-020-01542-z
Kajian sistematik pernah bermaksud kertas, pen dan kesabaranSystematic reviews used to mean paper, pens and patience
Cochrane Collaboration ditubuhkan, sempena nama Archie Cochrane yang menyeru ringkasan bukti yang teratur dan dikemas kini.The Cochrane Collaboration is founded, named after Archie Cochrane, who called for organised, up-to-date summaries of evidence.
Dua penilai membaca setiap abstrak bercetak, dan bertemu untuk menyelesaikan percanggahan.Two reviewers read every printed abstract, then met to resolve disagreements.
Alat menyusun dan mencadang; dua manusia masih membuat keputusan. Prinsipnya tidak berubah.Tools rank and suggest; two humans still decide. The principle hasn't changed.
"Gunung Abstrak""Abstract Mountain"
Pembelajaran aktif menyusun; LLM menilai. Kedua-duanya diaudit manusia.Active learning ranks; LLMs judge. Both are audited by humans.
Anda tetap membuat setiap keputusan; AI hanya mengubah urutan supaya rekod relevan muncul awal. Ketelusan tinggi, sesuai untuk tesis.You still make every decision; AI only changes the order so relevant records appear early. Highly transparent, thesis-friendly.
LLM membuat keputusan pertama dengan alasan. Pantas, tetapi mesti dikalibrasi pada sampel berlabel dan dilaporkan (alat, versi, tarikh).The LLM makes a first decision with reasons. Fast, but must be calibrated on a labelled sample and reported (tool, version, date).
van de Schoot, R., et al. (2021). An open source machine learning framework for efficient and transparent systematic reviews. Nature Machine Intelligence, 3, 125–133. doi:10.1038/s42256-020-00287-7 · Ouzzani, M., et al. (2016). Rayyan. Syst Rev, 5, 210. doi:10.1186/s13643-016-0384-4
Tiada model terbaik untuk semua; pilih ikut risiko yang anda sanggup tanggungNo model is best at everything; choose by the risk you can tolerate
Gemini 1.5 Pro & Claude 3.5 Sonnet cenderung sensitiviti; GPT-4o & Llama 3.3 cenderung spesifisiti.Gemini 1.5 Pro & Claude 3.5 Sonnet lean to sensitivity; GPT-4o & Llama 3.3 lean to specificity.
Untuk kajian sistematik, terlepas kajian lebih mahal daripada membaca lebih. Utamakan sensitiviti, kemudian manusia menapis positif palsu.For systematic reviews, missing a study costs more than reading extra. Prioritise sensitivity, then humans filter false positives.
Model berubah cepat; angka ini untuk versi 2024–25. Kalibrasi semula pada data anda sendiri.Models change fast; these figures are for 2024–25 versions. Recalibrate on your own data.
Oami, T., Okada, Y., & Nakada, T. (2025). Optimal large language models to screen citations for systematic reviews. Research Synthesis Methods, 16(6), 859–875. doi:10.1017/rsm.2025.10014
Minta keputusan dan sebab, dalam format yang boleh diauditAsk for a decision and a reason, in an auditable format
Anda penilai saringan kajian sistematik.You are a systematic-review screener.INKLUSI/INCLUSION: I1 belia/youth 15–30; I2 B40 / low income; I3 well-being outcome; I4 Malaysia.EKSKLUSI/EXCLUSION: E1 ulasan/editorial; E2 tiada data empirikal.RULE: Jika ragu-ragu, pilih INCLUDE.If uncertain, choose INCLUDE.OUTPUT: JSON {id, decision, criteria, reason_ms}[REKOD 1–3 di sini / RECORDS 1–3 here]
[ {"id":1,"decision":"INCLUDE","criteria":"I1,I2,I3,I4", "reason_ms":"Tinjauan 412 belia B40 di KL; ukur DASS-21."}, {"id":2,"decision":"EXCLUDE","criteria":"E1", "reason_ms":"Artikel ulasan naratif, tiada data primer."}, {"id":3,"decision":"INCLUDE","criteria":"I1,I3,I4 (I2?)", "reason_ms":"Status B40 tidak jelas; ragu → include."} ]
Rekod latihan rekaan. Rekod 3 menunjukkan peraturan "ragu → masukkan" berfungsi.Invented training records. Record 3 shows the "unsure → include" rule at work.
RULE: INCLUDE hanya jika SEMUA I1–I4 dinyatakan secara jelas dalam abstrak. Jika tidak, EXCLUDE dan nyatakan kriteria yang tiada.INCLUDE only if ALL of I1–I4 are explicit in the abstract. Otherwise EXCLUDE and name the missing criterion.
Kajian naratif atau skop dengan ribuan rekod, apabila masa terhad dan terlepas sedikit kajian boleh diterima. Rekod 3 kini akan EXCLUDE (I2 tiada): lihat kos ketegasan.Narrative or scoping reviews with thousands of records, when time is short and missing a few studies is acceptable. Record 3 now becomes EXCLUDE (I2 missing): the cost of strictness.
ID 4 · "Hubungan antara tekanan kos sara hidup dan kepuasan hidup dalam kalangan mahasiswa berpendapatan rendah di Selangor" · Kajian keratan rentas, n = 280, skala SWLS.Cross-sectional study, n = 280, SWLS scale.
{"id":4,"decision":"INCLUDE","criteria":"I1,I2,I3,I4", "reason_ms":"Mahasiswa (belia), berpendapatan rendah ≈ B40, SWLS = kepuasan hidup, Selangor."}
LLM memahami BM dengan baik, tetapi semak padanan istilah tempatan (contoh: "mahasiswa" = belia?). Keputusan definisi itu milik anda.LLMs handle Malay well, but check local term mapping (e.g. is "mahasiswa" = youth?). That definitional call is yours.
Tanya PDF anda sendiri, dan tuntut petikan untuk setiap jawapanAsk your own PDFs, and demand a quote for every answer
Bina jadual untuk setiap sumber: Lokasi | Reka bentuk | n | Alat ukur | Dapatan utama. Untuk setiap sel, sertakan petikan tepat dan nombor halaman. Tulis "tidak dinyatakan" jika tiada.Build a table for each source: Location | Design | n | Instrument | Key finding. For each cell, include an exact quote and page number. Write "not reported" if absent.
| Kajian latihanTraining study | LokasiLocation | Reka bentukDesign | n | Alat ukurInstrument | DapatanFinding | BuktiEvidence |
|---|---|---|---|---|---|---|
| A | Kuala Lumpur | Keratan rentasCross-sectional | 412 | DASS-21 | Tekanan kewangan ↑ → kemurungan ↑Financial stress ↑ → depression ↑ | "…β = 0.34…" p.7 |
| B | Kelantan | Kualitatif, temu bualQualitative, interviews | 18 | — | Sokongan keluarga sebagai penampanFamily support as a buffer | "…keluarga…" p.12 |
| C | Selangor | Keratan rentasCross-sectional | 280 | SWLS | Kos sara hidup ↓ kepuasan hidupCost of living ↓ life satisfaction | "…r = −0.29…" p.5 |
| D | Sabah | Kaedah campuranMixed methods | 156 | WEMWBS | Hutang tidak formal paling memberi tekananInformal debt most stressful | "…pinjaman…" p.9 |
| E | Pulau Pinang | LongitudinalLongitudinal | ⚠ tidak dinyatakannot reported | GHQ-12 | Kesan berkurang selepas BKMEffect weakens after cash aid | ⚠ tiada petikan · semak manualno quote · check manually |
Sel tanpa petikan ialah sel yang paling perlu anda baca sendiri. Data sebenar tidak dimuat naik ke alat awam tanpa kelulusan (Garis Panduan UKM 2.3).A cell without a quote is the cell you most need to read yourself. Never upload restricted data to public tools without approval (UKM Guideline 2.3).
Susun ikut konsep, bukan ikut pengarang, dan jurang akan kelihatanOrganise by concept, not by author, and the gaps appear
| KajianStudy | Tekanan kewanganFinancial stress | Sokongan sosialSocial support | Literasi kewanganFinancial literacy | Bantuan tunaiCash aid | Luar bandarRural | GenderGender |
|---|---|---|---|---|---|---|
| A | ● | ● | ||||
| B | ● | ● | ● | ● | ||
| C | ● | ● | ||||
| D | ● | ● | ● | |||
| E | ● | ● |
Bantuan tunai (1/5) dan gender (1/5) jarang dikaji. Tiada kajian menggabungkan literasi kewangan dengan luar bandar.Cash aid (1/5) and gender (1/5) are rarely studied. No study combines financial literacy with rural settings.
"Adakah literasi kewangan menyederhana kesan tekanan kewangan terhadap kesejahteraan belia B40 luar bandar?""Does financial literacy moderate the effect of financial stress on the well-being of rural B40 youth?"
Berdasarkan matriks ini sahaja, senaraikan gabungan konsep yang belum dikaji. Jangan tambah kajian luar. Cadangkan 2 soalan kajian untuk setiap jurang.Using only this matrix, list concept combinations not yet studied. Do not add outside studies. Suggest 2 research questions per gap.
Webster, J., & Watson, R. T. (2002). Analyzing the past to prepare for the future: Writing a literature review. MIS Quarterly, 26(2), xiii–xxiii. jstor.org/stable/4132319
AI mencadangkan kod; penyelidik membina temaAI proposes codes; the researcher builds themes
Ini lajur "Dapatan" daripada 23 kajian dalam matriks saya.Here is the "Findings" column from 23 studies in my matrix.Langkah 1: jana kod awal, setiap kod dengan ID kajian.Step 1: generate initial codes, each with study IDs.Langkah 2: kumpulkan kepada 3–5 tema calon, dengan bilangan kajian.Step 2: group into 3–5 candidate themes, with study counts.Langkah 3: senaraikan dapatan yang bercanggah dengan setiap tema.Step 3: list findings that contradict each theme.
Gunakan Teori Tekanan Keluarga (Family Stress Model) sebagai kerangka. Petakan setiap dapatan kepada komponen: tekanan ekonomi → tekanan emosi → konflik keluarga → kesejahteraan. Tandakan dapatan yang tidak sesuai dengan mana-mana komponen.Use the Family Stress Model as the framework. Map each finding to its components: economic pressure → emotional distress → family conflict → well-being. Flag findings that fit no component.
Sesuai apabila tesis anda berlandaskan teori tertentu. Dapatan yang "tidak muat" menjadi sumbangan teori anda.Fits when your thesis rests on a specific theory. Findings that "don't fit" become your theoretical contribution.
Bertindak sebagai penilai jurnal yang skeptikal. Untuk setiap tema saya, nyatakan (a) bukti yang lemah, (b) penjelasan alternatif, (c) kajian dalam senarai saya yang bercanggah. Jangan memuji.Act as a sceptical journal reviewer. For each of my themes, state (a) weak evidence, (b) alternative explanations, (c) studies in my list that contradict it. Do not praise.
Integriti: gunakan AI untuk mencabar tafsiran anda, bukan mengesahkannya.Integrity: use AI to challenge your interpretation, not to confirm it.
Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology, 3(2), 77–101. doi:10.1191/1478088706qp063oa · Thomas, J., & Harden, A. (2008). Thematic synthesis. BMC Med Res Methodol, 8, 45. doi:10.1186/1471-2288-8-45
Uji keputusan saringan dan analisis andaTest your screening and analysis judgement
Bina Artifak 03 + 04Build Artefacts 03 + 04
Tulis kriteria I/E bernombor. Saring abstrak dengan prompt Demo 4 (Gemini). Bandingkan dengan keputusan anda sendiri.Write numbered I/E criteria. Screen abstracts with the Demo 4 prompt (Gemini). Compare with your own decisions.
Isi nombor anda dalam carta PRISMA interaktif (slaid 2.1).Enter your numbers in the interactive PRISMA flow (slide 2.1).
Muat naik PDF akses terbuka ke NotebookLM (akaun UKM). Jalankan prompt Demo 5. Tandakan sel tanpa petikan.Upload open-access PDFs to NotebookLM (UKM account). Run the Demo 5 prompt. Flag cells without quotes.
Bina matriks konsep ringkas: 5 kajian × 4 konsep. Bulatkan satu jurang.Build a quick concept matrix: 5 studies × 4 concepts. Circle one gap.
Hanya muat naik artikel akses terbuka atau yang dibenarkan lesen PTSL.Only upload open-access articles or those your PTSL licence allows.
Automasi hujung-ke-hujung sudah tiba; piawaian pelaporan mengejarEnd-to-end automation has arrived; reporting standards are catching up
Kenyataan bersama 4 organisasiJoint statement, 4 organisations
Cochrane, Campbell, JBI & CEE: dedahkan AI yang membuat atau mencadangkan penilaian; lapor alat, versi, tarikh, pengesahan dan had.Cochrane, Campbell, JBI & CEE: disclose AI that makes or suggests judgements; report tool, version, date, validation and limits.
otto-SR
Sistem LLM hujung-ke-hujung: ketepatan ekstrak 93.1% berbanding 79.7% manusia. Pracetak: tunggu penilaian rakan.End-to-end LLM system: 93.1% extraction accuracy vs 79.7% for humans. Preprint: await peer review.
Elicit Systematic Review · PRISMA 2020
Aliran kerja ikut PRISMA 2020 dan boleh diaudit. Vendor melaporkan 95% dapatan semula carian dan 96% ketepatan ekstrak pada 994 ulasan Cochrane.PRISMA 2020-aligned, auditable workflow. Vendor reports 95% search recall and 96% extraction accuracy across 994 Cochrane reviews.
elicit.com/blog · dakwaan vendorvendor claim
LLM padat untuk saringanCompact LLMs for screening
Model kecil yang boleh dijalankan secara setempat sedang dinilai untuk saringan: penting untuk data sensitif yang tidak boleh keluar dari universiti.Small models that run locally are being evaluated for screening: important for sensitive data that cannot leave the university.
Implikasi: kemahiran bernilai beralih daripada melakukan saringan kepada mereka bentuk, mengkalibrasi dan melaporkan saringan.Implication: the valuable skill shifts from doing screening to designing, calibrating and reporting it.
TULIS & ETIKAWRITE & ETHICS
Dari matriks kepada hujah: menulis sintesis yang suara anda sendiri, dengan AI sebagai pengkritik dan pendedahan yang jujur.From matrix to argument: writing a synthesis in your own voice, with AI as critic and honest disclosure.
🎯 Artifak 05: perenggan sintesis + kenyataan pendedahan AI.🎯 Artefact 05: synthesis paragraph + AI disclosure statement.
Ringkasan berkata "siapa kata apa"; sintesis berkata "apa maknanya"A summary says "who said what"; a synthesis says "what it means"
Masalah: berpusat pengarang, tiada hubungan antara kajian, tiada jurang.Problem: author-centred, no links between studies, no gap.
Berpusat konsep · bukti digabung · perbandingan · jurang.Concept-centred · pooled evidence · comparison · gap.
Kajian A–E ialah data latihan rekaan dari Demo 5.Studies A–E are invented training data from Demo 5.
Gunakan AI sebagai pengkritik, bukan penulis hantuUse AI as a critic, not a ghostwriter
Anda pemeriksa luar tesis PhD sains sosial.You are an external examiner for a social-science PhD.JANGAN tulis semula perenggan saya.Do NOT rewrite my paragraph.Nilai: (1) dakwaan jelas? (2) bukti digabung? (3) perbandingan? (4) jurang?Assess: (1) clear claim? (2) pooled evidence? (3) comparison? (4) gap?Tandakan setiap ayat yang perlukan sitasi tetapi tiada.Flag every sentence that needs a citation but has none.Akhiri dengan 3 soalan yang pemeriksa akan tanya dalam viva.End with 3 questions an examiner would ask at the viva.
Anda penilai jurnal Q1 dalam bidang dasar sosial. Nilai bahagian kajian literatur saya dari segi: kebaharuan jurang, keadilan terhadap bukti bercanggah, dan kesesuaian dengan skop jurnal [nama jurnal]. Beri skor 1–5 dan keputusan: terima / pembetulan kecil / besar / tolak. Jangan tulis semula.You are a Q1 journal reviewer in social policy. Assess my literature review section for: novelty of the gap, fairness to contradicting evidence, and fit with the scope of [journal name]. Score 1–5 and decide: accept / minor / major / reject. Do not rewrite.
Tambah garis panduan penulis jurnal sebenar sebagai input supaya kritikan mengikut kriteria jurnal itu.Add the journal's real author guidelines as input so critique follows that journal's criteria.
Semak perenggan ini untuk Bahasa Melayu baku (rujuk Tatabahasa Dewan). Senaraikan kesilapan ejaan, imbuhan dan bahasa rojak dalam jadual: asal | cadangan | sebab. Jangan ubah istilah teknikal atau hujah.Check this paragraph for standard Malay (per Tatabahasa Dewan). List spelling, affix and code-mixing errors in a table: original | suggestion | reason. Do not change technical terms or the argument.
Penyuntingan bahasa ialah penggunaan dibenarkan (Garis Panduan UKM 2.2), tetapi setiap cadangan anda terima atau tolak sendiri.Language editing is a permitted use (UKM Guideline 2.2), but you accept or reject each suggestion yourself.
Enam kes mahkamah daripada ChatGPT. Semuanya rekaan.Six court cases from ChatGPT. All invented.
Dalam Mata lwn. Avianca (S.D.N.Y., Jun 2023), peguam menghantar hujahan yang menyitasi kes rekaan ChatGPT, dan didenda apabila tiada siapa dapat menemuinya.In Mata v. Avianca (S.D.N.Y., June 2023), lawyers filed a brief citing cases ChatGPT invented, and were fined when nobody could find them.
Mata v. Avianca, Inc. (Wikipedia) · Bloomberg Law, 22 Jun 2023
Boleh ✓ atau Tidak ✗? Ikut Garis Panduan GenAI UKM (2025)Allowed ✓ or Not ✗? Per the UKM GenAI Guideline (2025)
Garis Panduan Penggunaan Bertanggungjawab GenAI dalam Pengajaran, Pembelajaran dan Penyeliaan Penyelidikan UKM (2025), §2.2–2.3.UKM Guideline on Responsible Use of GenAI in Teaching, Learning, and Research Supervision (2025), §2.2–2.3. ukm.my · PDF · COPE (2023). Authorship and AI tools. publicationethics.org
Sebelum tampal atau muat naik, semak warna data andaBefore you paste or upload, check your data's colour
Hijau · bolehGreen · go
- Artikel akses terbuka (lesen CC)Open-access articles (CC licence)
- Teks yang anda tulis sendiriText you wrote yourself
- Rentetan carian, kriteria I/ESearch strings, I/E criteria
Kuning · semak dahuluAmber · check first
- Artikel berlesen langganan PTSL (semak terma penerbit)PTSL subscription articles (check publisher terms)
- Draf belum terbit milik rakanColleagues' unpublished drafts
- Bab tesis PhD: perlu kelulusan penyeliaPhD thesis chapters: needs supervisor approval
Merah · janganRed · stop
- Transkrip, rakaman, data respondenTranscripts, recordings, respondent data
- Maklumat peribadi boleh dikenal pastiPersonally identifiable information
- Dokumen sulit atau di bawah kelulusan etikaConfidential documents or ethics-restricted data
"Jika ragu, jangan muat naik." Guna versi institusi yang selamat (contoh: Gemini dalam Workspace UKM) dan patuhi skop kelulusan etika kajian anda."When uncertain, do not upload." Use secure institutional versions (e.g. Gemini in UKM Workspace) and stay within your study's ethics approval.
"Siapa Tulis Ini?""Who Wrote This?"
Bina kenyataan pendedahan yang lengkap dalam 60 saatBuild a complete disclosure statement in 60 seconds
Elemen mengikut Lampiran B2 Garis Panduan UKM dan kenyataan bersama Cochrane/Campbell/JBI/CEE (2025).Elements follow UKM Guideline Appendix B2 and the joint Cochrane/Campbell/JBI/CEE statement (2025).
Negara → universiti → jurnal: patuhi lapisan yang paling ketatNation → university → journal: follow the strictest layer
AIGE · 7 prinsipprinciples
Keadilan · Kebolehpercayaan, keselamatan & kawalan · Privasi & keselamatan data · Keterangkuman · Ketelusan · Akauntabiliti · Manfaat manusiaFairness · Reliability, safety & control · Privacy & data security · Inclusivity · Transparency · Accountability · Human benefit
Garis Panduan GenAIGenAI Guideline
PhD: kelulusan penyelia sebelum guna AI dalam bahagian material. Kelulusan untuk tugas penyelidikan tidak meliputi penulisan. Simpan prompt dan versi perantaraan.PhD: supervisor approval before using AI in any material part. Approval for research tasks does not cover writing. Keep prompts and intermediate versions.
AI bukan pengarangAI is not an author
Pengarang bertanggungjawab penuh dan mesti mendedahkan penggunaan AI. Lokasi pendedahan (Metodologi, Penghargaan) berbeza ikut jurnal: semak panduan penulis.Authors are fully responsible and must disclose AI use. Where to disclose (Methods, Acknowledgements) varies by journal: check author guidelines.
Jika jurnal membenarkan tetapi penyelia belum meluluskan, jangan guna dahulu. Lapisan paling ketat yang menentukan.If the journal allows it but your supervisor hasn't approved, don't use it yet. The strictest layer decides.
Penulisan dan etika: tiga senario sebenarWriting and ethics: three real scenarios
Peraturan menjadi lebih jelas; kerja penyelidikan menjadi lebih kolaboratifRules are getting clearer; research work is getting more collaborative
5 tahap AI dalam pentaksiran5 AI levels in assessment
Setiap pentaksiran mengisytiharkan tahap 1 (Tiada AI) hingga 5 (Penerokaan AI); kursus memilih Initiation, Integration atau Infusion.Each assessment declares level 1 (No AI) to 5 (AI Exploration); courses choose Initiation, Integration or Infusion.
Pengesan AI bukan buktiDetectors are not proof
Skor pengesan adalah kebarangkalian; tiada keputusan akademik buruk semata-mata berdasarkannya. Simpan prompt dan versi draf sebagai bukti proses.Detector scores are probabilistic; no adverse decision rests on them alone. Keep prompts and drafts as process evidence.
Penyelidikan "berbilang pemain""Multiplayer" research
Alat seperti Elicit kini menyokong pustaka dan sesi penyelidikan berkongsi: kumpulan penyeliaan boleh membina asas bukti bersama. Konektivisme dalam produk.Tools like Elicit now support shared libraries and research sessions: supervision groups can build an evidence base together. Connectivism as a product.
Ejen lebih banyak sitasi, lebih banyak risikoMore citations, more risk
Ejen deep research menghasilkan lebih banyak sitasi setiap pertanyaan daripada LLM berbantu carian, tetapi kadar URL rekaan lebih tinggi.Deep-research agents produce more citations per query than search-augmented LLMs, but hallucinate URLs at higher rates.
Implikasi: dokumentasi proses (prompt, versi, log keputusan) kini sama penting dengan hasil akhir.Implication: process documentation (prompts, versions, decision logs) now matters as much as the final product.
Tandakan artifak yang anda sudah milikiTick the artefacts you now have
Yang belum siap menjadi tugasan 7 hari anda. Pulang dengan proses, bukan hanya alat.Anything unfinished is your 7-day task. Leave with a process, not just tools.
Alat akan berubah; rangkaian anda yang mengekalkan kemahiranTools will change; your network keeps you skilled
Lengkapkan artifak yang belum siap. Simpan semua prompt dalam satu dokumen log.Finish incomplete artefacts. Keep all prompts in one log document.
Bincang skop penggunaan AI dengan penyelia; dapatkan kelulusan bertulis jika PhD.Discuss AI-use scope with your supervisor; get written approval if PhD.
Bina kumpulan Zotero bersama rakan bengkel; kongsi satu rentetan carian dan satu jurang setiap minggu.Start a shared Zotero group with workshop peers; share one search string and one gap each week.
Pek prompt: 0 prompt bengkel, sedia digunakanPrompt pack: every workshop prompt, ready to use
Butang ⧉ + ↗ menyalin prompt dan membuka alatnya dalam tab baharu: tampal dengan Ctrl+V. Tekan C pada mana-mana slaid untuk menyalin prompt pertamanya.The ⧉ + ↗ button copies the prompt and opens its tool in a new tab: paste with Ctrl+V. Press C on any slide to copy its first prompt.
Sumber & bacaan lanjutResources & further reading
Kredit foto & ilustrasiPhoto & illustration credits
Terima kasihThank you
AI mempercepat pencarian. Integriti memastikan apa yang kita temui layak dipercayai.AI speeds up the search. Integrity makes what we find worth trusting.
Terima kasih kepada Sekretariat Penyelidikan & Inovasi FSSK dan Perpustakaan Tun Seri Lanang.With thanks to the FSSK Research & Innovation Secretariat and Tun Seri Lanang Library.