2023 IB DP Computer Science Practice Paper with Answers

Beta testing involves releasing a pre-release version to a select group of target users in real-world environments. Without it, diverse system compatibilities and edge cases remain untested. This can lead to critical bugs post-release, resulting in negative user feedback, increased customer support demands, and the urgent need to deploy expensive, emergency hotfixes.

Award [1 mark] for identifying a valid consequence (e.g., reputational damage, high support costs, critical bugs). Award [1 mark] for explaining the mechanism (e.g., because real-world user environments are highly diverse and unsimulated in developer environments). Award [0.7 marks] for linking this directly to the organizational impact (e.g., loss of revenue or high cost of emergency patch cycles).

During the Fetch phase, the CPU copies the memory address stored in the Program Counter (PC) to the Memory Address Register (MAR). The PC is then automatically incremented by the control unit to point to the address of the next instruction in sequence, ensuring continuous execution of the program.

Award [1 mark] for stating that the PC holds or stores the address of the next instruction to be executed. Award [1 mark] for explaining that its value is copied or transferred to the Memory Address Register (MAR) during the fetch stage. Award [0.7 marks] for noting that it increments or updates automatically to point to the subsequent instruction.

Without a common protocol, receiving devices would not know how to interpret incoming electrical, optical, or radio signals, how to extract data packets, how to perform error checking, or how to handle flow control. Standardized rules guarantee compatibility and error-free delivery between diverse hardware and software architectures.

Award [1 mark] for defining a protocol as a standardized set of rules. Award [1 mark] for explaining that it ensures interoperability or compatibility between different hardware and software configurations. Award [0.7 marks] for mentioning a specific function such as error checking, flow control, or packet formatting.

Bubble sort works by continuous adjacent comparisons and swaps throughout each pass, gradually moving the largest or smallest elements to the end of the list. In contrast, selection sort performs multiple comparisons to locate the minimum or maximum value in the unsorted portion of the list, but only performs a single swap at the end of each pass once the target element is identified.

Award [1 mark] for describing bubble sort's mechanism (adjacent comparisons and multiple swaps per pass). Award [1 mark] for describing selection sort's mechanism (scanning for min/max and swapping once per pass). Award [0.7 marks] for explicitly contrasting the frequency of swaps or the focus of the comparisons (adjacent vs global scan).

A dynamic linked list allocates memory as nodes are added, meaning there is no fixed upper limit, unlike a static array which has a predetermined size. However, linked lists do not support direct indexing (meaning access by index is \(O(n)\) instead of \(O(1)\) for arrays) and require additional memory for storing pointers in each node.

Award [1 mark] for a valid advantage (e.g., dynamic size, efficient insertions/deletions). Award [1 mark] for a valid disadvantage (e.g., pointer overhead, lack of random access / sequential search required). Award [0.7 marks] for linking these properties back to performance or resource utilization.

When physical RAM is exhausted, the operating system allocates space on secondary storage (such as a hard disk or solid-state drive) known as swap space. Through paging, inactive memory blocks are transferred to the disk, allowing active processes to run. If the swapped-out data is requested again, a page fault occurs, prompting the OS to swap out other data to bring the requested data back into physical RAM.

Award [1 mark] for explaining that virtual memory uses secondary storage to simulate primary memory. Award [1 mark] for describing the process of paging or swapping inactive memory blocks from RAM to the storage drive. Award [0.7 marks] for explaining that this allows programs larger than physical memory to execute, despite slower processing speeds.

A sensor functions as an input device, continually monitoring physical characteristics of the environment (such as a temperature sensor measuring the greenhouse air temperature) and sending data to the processor. An actuator functions as an output device, receiving commands from the central processor and converting electrical energy into physical action (such as starting a motor to lift open the ventilation panels).

Award [1 mark] for defining the role of the sensor as an input/measurement device transforming physical stimuli into data. Award [1 mark] for defining the role of the actuator as an output/action device transforming data commands into physical movement. Award [0.7 marks] for contextualizing both within the greenhouse ventilation system.

Direct changeover replaces the old system instantly with the new one. If the new system fails, there is no backup, representing high risk, but it avoids running duplicate systems. Phased introduction implements the new system in parts (e.g., one module or department at a time), which allows issues to be isolated and resolved with lower overall risk, although it requires running both systems concurrently for a period.

Award [1 mark] for describing direct changeover and its characteristics (instantaneous, high risk, fast/inexpensive). Award [1 mark] for describing phased introduction and its characteristics (modular rollout, lower risk, slower/complex). Award [0.7 marks] for providing a clear comparative point, such as contrasting the level of operational risk against implementation cost or time.

A router performs three key functions: First, it receives incoming packets and inspects their headers to extract the destination IP address. Second, it consults its routing table, which contains network topology information, to determine the most efficient next hop or network interface for the packet based on metrics like distance or congestion. Third, it forwards the packet from the incoming interface to the appropriate outgoing interface, sending it closer to its final destination.

Award [1] mark for each distinct role identified, up to a maximum of [3] marks: [1 mark] for reading or inspecting the packet header to find the destination IP; [1 mark] for consulting a routing table or applying algorithms to determine the optimal path or next hop; [1 mark] for forwarding the packet to the appropriate output interface or next node.

(a) Stakeholders: Doctors, nurses, administrative clerks, IT support technicians, or hospital management. (b) Advantage of direct changeover: The changeover is rapid and immediate, which reduces operational costs of maintaining dual systems. Disadvantage: If the new system fails, there is no backup system, which could lead to critical disruption of patient care. (c) Data loss/corruption: Can occur due to mismatched database schemas, format incompatibilities, or transfer interruption. Mitigation: Create complete system backups before migration and perform validation checks like database checksums. (d) UAT involves representative healthcare users performing daily operational tasks. This helps reveal non-intuitive controls, workflow bottlenecks, and errors in medical record entries before system launch. (e) Prototypes are built to gather early user feedback and to verify design feasibility.

(a) [2 marks] Award 1 mark for each identified stakeholder up to 2. (b) [4 marks] Award 2 marks for advantage (1 mark for statement, 1 mark for explanation) and 2 marks for disadvantage (1 mark for statement, 1 mark for explanation). (c) [3 marks] Award 1 mark for identifying a source of data loss/corruption, and 2 marks for a well-described mitigation measure. (d) [4 marks] Award 1 mark per point explaining UAT benefits: identifies real-world interface problems (1), reduces training curve (1), ensures functional alignment (1), gathers genuine feedback from non-technical end-users (1). (e) [2 marks] Award 1 mark per valid reason up to 2: visual representation for feedback, cost-saving detection of flaws, or verifying feasibility.

(a) Packet switching is a networking method where data is split into packets that are routed independently across optimal paths and then reassembled at the destination. (b) TCP divides data into packets with sequence numbers, performs a handshake to establish a connection, and requires the receiver to send an acknowledgment (ACK). Missing packets trigger a timeout and retransmission. IP handles routing. (c) (i) Factor 1: Metal structures causing signal absorption or reflection. Factor 2: Electromagnetic interference (EMI) from heavy electric motors. (ii) Risk: Interception of data because wireless signals travel through open space. Mitigation: Use WPA3 encryption. (d) A router forwards data packets between different networks by reading destination IP addresses and selecting the best path.

(a) [2 marks] Award 1 mark for breaking data into packets, 1 mark for independent routing/reassembly. (b) [4 marks] Award 1 mark for sequence numbers, 1 mark for handshake/connection, 1 mark for receiver acknowledgments (ACK), 1 mark for timeout and retransmission of lost packets. (c)(i) [4 marks] Award 2 marks per factor: 1 mark for identifying factor, 1 mark for explanation. (c)(ii) [3 marks] Award 1 mark for security risk (interception), 1 mark for explanation, 1 mark for naming protocol (WPA2/WPA3). (d) [2 marks] Award 1 mark for connecting networks, 1 mark for routing decisions using routing tables/IP addresses.

(a) Algorithm: MAX_LEN = 0; CUR_LEN = 0; MAX_START = -1; loop I from 0 to 719; if TEMPS[I] < 0.0 then; if CUR_LEN == 0 then CUR_START = I; end if; CUR_LEN = CUR_LEN + 1; if CUR_LEN > MAX_LEN then MAX_LEN = CUR_LEN; MAX_START = CUR_START; end if; else; CUR_LEN = 0; end if; end loop; output 'Length: ', MAX_LEN; output 'Index: ', MAX_START. (b) Modularity breaks code into independent, manageable modules, enhancing readability, allowing team development, simplifying debugging, and facilitating code reuse. (c) Average-case: O(N). Worst-case: O(N). Justification: If the element is at the very end or not in the array, all N elements must be searched, which scales linearly with size.

(a) [8 marks] Award marks as follows: - Initialize tracking variables (1 mark); - Loop through all 720 elements (1 mark); - Correct condition check for below freezing (1 mark); - Set start index of current sequence correctly (1 mark); - Correctly increment current sequence length (1 mark); - Compare current length with max length and update values only if strictly greater (2 marks); - Reset current length to 0 on non-freezing temperature (1 mark). (b) [3 marks] Award 1 mark per benefit up to 3: modular isolation simplifies testing, reusable parts save time, team division of work, clearer design logic. (c) [4 marks] Award 1 mark for average case O(N), 1 mark for worst case O(N), and 2 marks for explaining that worst-case requires looking at all N elements.

(a) A static queue has a fixed size allocated at compile time. A dynamic queue grows and shrinks at runtime using pointers. Advantage of static: lower memory overhead and execution speed. Advantage of dynamic: prevents overflow errors and manages memory efficiently. (b) Pseudocode: method enqueue(ticketID); if count == 10 then; output 'Overflow'; else; if count == 0 then; head = 0; tail = 0; else; tail = (tail + 1) mod 10; end if; TICKETS[tail] = ticketID; count = count + 1; end if; end method. (c) A queue is linear, sequential, and follows First-In-First-Out (FIFO). A BST is a hierarchical, sorted tree structure. Searching in a queue takes O(N) as elements must be scanned linearly. Searching in a BST takes average O(log N) because half the branches are pruned at each step.

(a) [4 marks] Award 1 mark for static/dynamic difference, 1 mark for static advantage, 1 mark for dynamic advantage, 1 mark for clear elaboration. (b) [6 marks] Award 1 mark for checking overflow (count == 10), 1 mark for handling initial empty queue (count == 0), 2 marks for modulo calculation of tail, 1 mark for writing element to array, 1 mark for incrementing count. (c) [5 marks] Award 1 mark for queue linear/FIFO explanation, 1 mark for BST hierarchical/sorted explanation, 1 mark for identifying search scenario, 2 marks for comparing search complexities (O(N) vs O(log N)).

(a) Sensor: measures physical soil moisture levels and sends this raw data as an input signal to the microprocessor. Actuator: takes output control signals from the microprocessor and performs physical work (opening or closing the water pump valve). (b) Open-loop system operates on a preset schedule (e.g. watering 10 minutes daily) without checking soil conditions. Closed-loop system continually measures soil moisture and uses it as feedback to decide whether to water. Closed-loop is superior because it conserves water and prevents plant damage. (c) Microprocessor steps: 1. Read input signal from the sensor. 2. Compare reading to target of 60%. 3. If reading is under 60%, activate the water pump actuator. 4. If reading is 60% or higher, deactivate/keep off the pump. 5. Wait/delay and repeat. (d) Considerations: water conservation/sustainability, potential job losses for manual farmworkers, and reliability (e.g., system failure causing crop death).

(a) [4 marks] Award 2 marks for sensor (1 mark for measuring physical environment, 1 mark for sending input signal). Award 2 marks for actuator (1 mark for receiving signal, 1 mark for executing mechanical action). (b) [4 marks] Award 1 mark for open-loop description, 1 mark for closed-loop feedback description, 2 marks for explaining why closed-loop is superior (avoids overwatering/underwatering based on external weather). (c) [4 marks] Award 1 mark for each step up to 4: reads sensor data, compares with 60% threshold, outputs control decision to actuator, and implements loop delay before repeating. (d) [3 marks] Award 1 mark for each valid point up to 3: environmental impact of water use, social impact on manual jobs, safety/dependability of system logic.

A hybrid recommender system effectively resolves the cold-start problem which occurs when a new farmer joins with zero historical data. In collaborative filtering, recommendations depend on matching the target user's historical ratings with similar users. Without historical data, this is impossible. Content-based filtering solves this initial stage by prompting the new user for static data during registration, such as regional zip code, average soil pH, and climate zone. The system matches these attributes to crop requirements to generate baseline recommendations. As the farmer inputs real performance data over subsequent seasons, collaborative filtering algorithms (such as K-Nearest Neighbors or Matrix Factorization) gradually take over, blending the two recommendation scores. This hybrid approach ensures the user receives relevant, localized suggestions from day one while benefiting from community-driven collaborative insights over time.

Maximum 7.5 marks total. Award up to 2 marks for a clear definition of the cold-start problem in the context of agricultural recommender systems (lack of historical crop yield or rating data for new users). Award up to 2 marks for explaining how content-based filtering operates using static registration metadata (climate, soil attributes, regional profile). Award up to 2 marks for demonstrating how the hybrid model integrates both approaches (e.g., weighted combination, switching mechanism). Award up to 1.5 marks for describing the transition dynamic (increasing collaborative weight as user interaction data accumulates over agricultural seasons).

Matrix Factorization, such as SVD, handles sparsity by projecting both farmers and crops into a shared latent factor space of dimension \(k\). The original sparse rating matrix \(R\) is approximated by the product of two dense matrices: \(U\) (farmer-latent factors) and \(V^T\) (crop-latent factors), such that \(R \approx U \times V^T\). This allows the system to predict a missing value by calculating the dot product of a farmer's latent vector and a crop's latent vector. Evaluation: (1) Advantages: It effectively uncovers hidden patterns (e.g., complex interactions between soil tolerance and climate) that are not explicitly labeled, and it scales better than pure neighborhood-based methods as sparsity increases. (2) Disadvantages: High computational cost to retrain the latent factor matrices when new data is added, meaning real-time updates are difficult. Additionally, the latent factors are mathematical abstractions, making the resulting recommendations a "black box" which reduces explainability for farmers who want to know why a crop was recommended.

Maximum 7.5 marks total. Award up to 2 marks for explaining the mechanism of Matrix Factorization (decomposition into lower-dimensional latent factor matrices representing farmers and crops). Award up to 2 marks for explaining how SVD resolves the sparsity issue (by mathematically reconstructing missing entries using the dot product of latent vectors). Award up to 2 marks for identifying technical limitations/trade-offs (high computational retraining cost, lack of real-time update capability, and lack of model explainability/transparency). Award up to 1.5 marks for structured synthesis and critical evaluation of its suitability for agricultural recommendations.

A major ethical and systemic risk of automated recommender systems in agriculture is the feedback loop, analogous to the digital "filter bubble." If the recommender system identifies that Crop A produces high yields in Region X, it will recommend Crop A to all nearby farmers. As more farmers plant Crop A, the algorithm receives more positive feedback data, further reinforcing its recommendation. Over several seasons, this leads to: (1) Ecological Degradation: Continuous farming of a single crop (monoculture) depletes specific soil nutrients, ruins local biodiversity, and increases vulnerability to devastating pest outbreaks. (2) Market Saturation: Simultaneous overproduction of a single crop by all regional farmers drives down market prices, resulting in economic hardship. (3) Loss of Indigenous Knowledge: Farmers may lose traditional crop rotation expertise and rely completely on a black-box proprietary algorithm, leading to technological dependency. Mitigations include incorporating exploration metrics (serendipity) in the loss function to intentionally recommend diverse, non-optimal but ecologically beneficial crop rotations.

Maximum 7.5 marks total. Award up to 2 marks for explaining the feedback loop mechanism in agricultural recommendations (positive reinforcement of dominant suggestions). Award up to 2 marks for analyzing ecological consequences (monoculture development, rapid soil nutrient depletion, pest vulnerability). Award up to 2 marks for explaining socioeconomic/ethical issues (market over-saturation, price depreciation, and tech dependency/loss of traditional local knowledge). Award up to 1.5 marks for proposing balanced algorithmic mitigations (e.g., adding serendipity/diversity constraints, forcing crop rotation logic into the recommendation engine).

Integrating real-time IoT soil data presents a structural architectural choice: (1) Edge-Computing Architecture: Sensor data is processed locally on a farm gateway device. Pros: Immediate latency-free inference, works completely offline which is critical for rural areas with poor connectivity, and localized data privacy is maintained. Cons: Gateways have limited computational resources (CPU, RAM, storage), meaning they cannot run heavy deep-learning model training locally; they can only perform lightweight model inference. (2) Centralized Cloud-Computing Architecture: Data is transmitted via cellular/satellite networks to centralized servers. Pros: Massive computational scale, allows continuous retraining of the recommendation model using data aggregated from thousands of farms, and facilitates global trend analysis. Cons: Highly dependent on continuous network connectivity; if connection fails, real-time alerts or recommendations cannot be computed. High data transmission costs. A hybrid approach (fog computing) is often optimal, where the edge performs immediate anomaly detection/inference while the cloud handles periodic model updates.

Maximum 7.5 marks total. Award up to 2 marks for analyzing the edge-computing option (highlighting benefits of offline execution, low latency, and constraints of local hardware power). Award up to 2 marks for analyzing the centralized cloud-computing option (highlighting benefits of massive scale, global data aggregation, and constraints of rural connectivity/bandwidth). Award up to 2 marks for comparing how each choice impacts dynamic recommender models specifically (local lightweight inference vs heavy centralized retraining). Award up to 1.5 marks for proposing and justifying a sensible hybrid/compromise architecture (e.g., Edge for inference, Cloud for training).