Games Researcher: Player Behavior, Retention, and Human-AI Interaction in Interactive Systems

I use behavioral data from systems that are actually in use to understand how people interact with adaptive and AI-driven software, mostly in games and interactive media, where I read player behavior and retention data to see how a design shapes what people do. My FDG 2026 paper is one example: across 35 mobile game prototypes, how often players returned predicted Day 1 retention better than how long they played, and expert coding of design qualities such as creative confidence added predictive value the automated metrics missed. I've run the same kind of evaluation on AI systems I built, measuring how a medical-report assistant changed users' clarity, confidence, and trust. I want to read behavior to see how a system shapes the decisions people make with it, without taking the decision away from them.

Rakib Jahan Siddique

Currently applying to PhD and research-master's programs in HCI, games research, and Human-AI interaction for a 2027 start. Open to research collaboration in the meantime. IELTS Academic 8.0 (CEFR C1).

In place of a master's: eight years of shipped R&D and a first-author FDG paper. Details under Academic Background.

Research

PUBLISHED2026

Session Architecture Predicts Mobile Game Retention: A Single-Studio Case Study

FDG '26: Proceedings of the 21st International Conference on the Foundations of Digital Games

Player BehaviorRetention & EngagementGames User ResearchA/B Testing & Evaluation

Looked at 35 mobile game prototypes from 2023 to 2025 to find which early behaviors predict Day 1 retention. Session frequency was the strongest predictor (r = 0.84), ahead of total playtime. Adding expert-coded design features, especially creative confidence (r = 0.63), added something beyond the automated metrics. Retention went up 71% over the 2.5-year window, alongside gains in design quality.

Motivation: Studios track a lot of engagement metrics, but which early signal actually tells you whether a player comes back? This study asks that across 35 prototypes with real analytics data.
Key Finding: Getting players back three times matters more than keeping them for 60 minutes. Session frequency, not total playtime, is the main retention signal.
PREPRINT2025

Medical Reports AI Assistant

Human-AI InteractionTrust & ExplainabilityRAG / RetrievalA/B Testing & Evaluation

A retrieval-based assistant that lets patients upload a medical report and query it in natural language, with every answer grounded in and citing the specific source passages it is drawn from. Implemented in Python with LangChain and the OpenAI API for parsing and summarization, plus OCR for scanned PDFs. Evaluated in a preregistered within-subject pilot (N = 5, non-medical participants) contrasting a manual search condition with the assistant across two sessions, measuring time-to-find, ease, and confidence.

Motivation: Patients face a comprehension barrier when reading their own reports: results are scattered across dense numerical tables and clinical terminology. The study asks whether a retrieval assistant can lower that barrier without eroding trust.
Key Finding: Time-to-find dropped 62%; self-reported ease rose from 3.2 to 4.4 and confidence from 3.0 to 4.0 on five-point scales. Adoption tracked trust design, particularly visible source citations, more than model capability. Upload latency and the absence of explicit accuracy guarantees remained limitations.
PREPRINT2026

Temporal Metrics and Perceived Fluency in L2 Speech: A Production-Based Analysis

NLP / SpeechA/B Testing & Evaluation

Tests whether a duration-weighted disfluency metric matches human fluency judgments better than the usual rate- or count-based measures in L2 speech. It looks at how fluent speech sounds, not overall speaking proficiency.

Research Interests

Games & Player Behavior

How game systems shape what players do: session patterns, retention signals, and design qualities read from behavioral data.

Human-AI Interaction

How people build trust and get a clear picture of what an agent is doing when they work with it.

Agentic AI

Agent workflows for multi-step reasoning and tool use, built around evaluation rather than vibes.

Featured Projects

Data-Driven Evaluation of Player Behavior in a Mobile Shooter

Mobile game - 100k+ downloads

Player BehaviorRetention & EngagementA/B Testing & EvaluationGames User Research
A/B testingRetention analytics100k+ downloads

"Tuning features against the analytics lifted D1 retention 10% and playtime 22%."

Music VR Project

Audio-reactive virtual environment

VR / XRHCIGame Design & Systems
XRUnityAudio-reactive

"Letting the environment react to sound, not just play it, makes a VR space feel alive."

Surfing VR Game with Mechanical Surfboard

Physical peripheral driving in-game navigation

VR / XRHCIGame Design & Systems
XRUnityHardware input

"A physical controller you move with your body pulls you into a VR scene in a way a handset doesn't."

Tuned Qwen 4B vs. frontier model benchmark

Fine-Tuning vs. Prompt Engineering for Task-Specific Models

When customizing model weights beats prompting

Applied ML / Fine-tuningNLP / SpeechA/B Testing & Evaluation
LoRA fine-tuneQwen3-4B0.915 macro F1

"A 4B model on a gaming GPU beat a frontier model ~100x its size on this task, on accuracy and cost."

Read
Puzzle game prototypes built with Voodoo

Brutal Learnings from Puzzle Games

Prototyping puzzle mechanics with Voodoo

Game Design & SystemsPlayer BehaviorRetention & Engagement
PrototypingVoodoo28% D1 best case

"Puzzles live or die on rule clarity and input responsiveness. Fusing two mechanics that fit beats bolting one onto another."

Grammar Sorts Demo

Applied Educational Game Systems

Learning & Behavior Research Through Interactive Design

Educational GamesGame Design & SystemsPlayer Behavior
Learning games4 prototypesUnity

"How a game is built changes what players learn: when feedback lands shapes strategy, difficulty pacing sets the mental load, and the choices you offer steer behavior."

Speech Analysis API (IELTS-Aligned)

English proficiency assessment system

NLP / SpeechApplied ML / Fine-tuning
AIWhisper / Wav2VecLive paid API

"After Whisper transcription and wav2vec analysis, it generates a score deterministically across the four rubrics, giving a reliable measurement of improvement."

App Store
LLM-driven detective game prototype

Turning Dialogue into Deduction

LLM-driven detective game prototype

Agentic AI / LLM AgentsGame Design & SystemsHuman-AI Interaction
LLM gameplayGeminiPrototype

"Structured AI dialogue can drive reasoning-based gameplay, pushing players to think instead of following a branch."

Read
Catching the Cow gameplay

Catching the Cow: Designing From a Cultural Moment

Runner game built from an Eid-ul-Adha moment

Game Design & SystemsPlayer Behavior

"Culture can replace complexity. When the context is shared, design can stay minimal and still feel rich."

Read

Methods & Technical Experience

AI AND AGENTIC SYSTEMS

  • OpenAI API
  • LangGraph / LangChain
  • AutoGen
  • Claude API
  • RAG pipelines
  • n8n Automation

EVALUATION & BEHAVIORAL ANALYSIS

  • A/B Testing
  • Retention Analytics (D1-D14)
  • Statistical Analysis
  • Behavioral Coding & Metrics

INTERACTIVE SYSTEMS & GAME DEV

  • Unity (C#)
  • Gameplay Architecture
  • GameAnalytics
  • Human-Computer Interaction

BACKEND & DATA SYSTEMS

  • FastAPI
  • Next.js
  • Supabase
  • Supabase VectorDB
  • BackBlaze B2

Academic Background

B.Sc. in Computer Science & Cognitive Systems

Islamic University of Technology, Bangladesh

2015 to 2018

CGPA: 3.57 / 4.00 · Final-two-year GPA: 3.74 / 4.00

Relevant Coursework: Human-Computer Interaction, Machine Learning, Computer Graphics & Multimedia Systems, Artificial Intelligence, Pattern Recognition, Cognitive Psychology, Computer Vision, Statistical & Mathematical Analysis

I hold a BSc rather than a master's. In place of one, I bring eight years of applied R&D across five game studios, running A/B tests, retention analysis, and player-behavior studies on shipped systems, and a first-author FDG 2026 paper. I read this as an equivalent research trajectory and am prepared to discuss direct PhD eligibility with prospective supervisors.

Additional Training

Game Design & Gamification

Metropolia University of Applied Sciences, Finland · 10 ECTS

2021

English Proficiency

IELTS Academic · Overall 8.0 (CEFR C1) · Listening 8.5, Reading 8.5, Speaking 7.5, Writing 7.0 · September 2025

Professional Affiliations

Member, International Game Developers Association (IGDA) · Volunteer, Events Committee

Professional Experience

Digital Games Developer

Gameloops, Dhaka, Bangladesh

Aug 2022 to present

Led mobile games and interactive systems from prototype to release. Ran A/B tests on FTUE, economy balance, and which features to keep. Used retention (D1/D7/D14), engagement, and revenue data to decide what to change next. Built and tested agentic AI workflows for automation and internal tools.

Digital Games Developer

Lightneer, Helsinki, Finland

Aug 2020 to Jul 2022

Built gameplay prototypes and revised them from player testing and UX feedback.

Digital Games Developer

KobGames Studios, Tel Aviv, Israel

Jan 2020 to Jun 2020

Designed and built gameplay systems across several prototypes.

Digital Games Developer

Alienide Interactive, Dhaka, Bangladesh

Sep 2018 to Dec 2019

Developed mobile applications, games, and AR/VR learning tools.

Applied Engineering

Client and product automation work and game design write-ups built and shipped outside of a research context, included for completeness, not as evaluated research.

Medical Reports AI Workbench

Medical Reports AI Workbench

Upload a report, ask questions, get cited answers

Human-AI InteractionRAG / RetrievalTrust & Explainability

Patients face a comprehension barrier when reading their own medical reports: results are scattered across dense numerical tables and clinical terminology. This work asks whether a retrieval-based assistant can lower that barrier without eroding trust.

Constraining AI Agent Behavior for Predictable Outcomes

Constraining AI Agent Behavior for Predictable Outcomes

Deterministic workflow with an AI interruption layer

Agentic AI / LLM AgentsAutomation / Workflow SystemsHuman-AI Interaction

Visual workflows break when users go off-script (changing passenger count mid-confirmation); pure AI agents handle that but are unpredictable for processes that need to follow rules. This tests a hybrid of the two.

AI Phone Receptionist with Shopify Checkout

AI Phone Receptionist with Shopify Checkout

Inbound calls to checkout links, no staff

Agentic AI / LLM AgentsAutomation / Workflow Systems

An AI receptionist that does more than answer questions: take a phone call, recommend a product, and get a checkout link into the customer's hands end to end.

AI Property Maintenance Coordinator

AI Property Maintenance Coordinator

Tenant report to contractor approval

Agentic AI / LLM AgentsAutomation / Workflow Systems

One maintenance report spins off a lot of coordination for a property manager: judging urgency, chasing details, finding contractors, checking availability, comparing quotes, getting approval. This automates the repetitive parts.

AI Book Translator

AI Book Translator

Multi-language document translation system

Applied ML / Fine-tuningAutomation / Workflow SystemsNLP / Speech

A translation pipeline that keeps meaning intact and checks its own quality across several languages.

WhatsApp AI Customer Support System

WhatsApp AI Customer Support System

Automated customer service chatbot

Agentic AI / LLM AgentsAutomation / Workflow Systems

Can a chatbot handle customer support, sort incoming issues, and write useful replies?

Mafia Marksman: Thematic Progression in a Mobile Shooter

Mafia Marksman: Thematic Progression in a Mobile Shooter

Mafia-rank progression driving retention

Game Design & SystemsRetention & EngagementPlayer Behavior

After Auto Cop's retention fell despite a rising CPI, this tested whether a mafia theme with intuitive, reward-linked progression and varied weapons could hold players and improve monetization.

Auto Cop: Three Iterations on a Cover Shooter

Auto Cop: Three Iterations on a Cover Shooter

CPI vs. retention, a roguelite detour, and back to core

Game Design & SystemsRetention & EngagementPlayer Behavior

Publishers were seeing strong returns from cover shooters, so the team built one aiming for 200% ROAS. This three-part case study follows what happened when strong install numbers met weak retention.

Three Ways to Onboard Players in Mobile Games

Three Ways to Onboard Players in Mobile Games

Teaching mechanics without breaking flow

Game Design & SystemsA/B Testing & EvaluationPlayer Behavior

Developers need to guide players through new mechanics without disrupting immersion. This looks at how to teach rules while keeping cognitive load low.

Adding Meaningful Gameplay to an Arcade Idle Game

Adding Meaningful Gameplay to an Arcade Idle Game

Mapping resource flows with Machinations

Game Design & Systems

Arcade idle games are resource-flow engines. This asks where meaningful player choice can be added without breaking hypercasual accessibility.

Slice Disc: Path-Slicing Mechanics Case Study

Slice Disc: Path-Slicing Mechanics Case Study

First-person slicing meets bouncing-disc physics

Game Design & SystemsRetention & Engagement

First-person slicing games were gaining traction in hypercasual, and third-party slicing assets promised strong CPI. This tested whether a slicing plus bouncing-disc loop could scale.