Massey Documents by Type

Permanent URI for this communityhttps://mro.massey.ac.nz/handle/10179/294

Browse

Filters

2020 - 20255

Settings

Start date: 2020Subject: search.filters.subject.Artificial intelligence

Search Results

Now showing 1 - 5 of 5
  • Thumbnail Image
    Item
    A Hormetic Approach to the Value-Loading Problem: Preventing the Paperclip Apocalypse
    (Springer Nature Singapore Pte Ltd, 2025-10-06) Henry NIN; Pedersen M; Williams M; Martin JLB; Donkin L
    The value-loading problem is a major obstacle to creating Artificial Intelligence (AI) systems that align with human values and preferences. Central to this problem is the establishment of safe limits for repeatable AI behaviors. We introduce hormetic alignment, a paradigm to regulate the behavioral patterns of AI, grounded in the concept of hormesis, where low frequencies or repetitions of a behavior have beneficial effects, while high frequencies or repetitions are harmful. By modeling behaviors as allostatic opponent processes, we can use either Behavioral Frequency Response Analysis (BFRA) or Behavioral Count Response Analysis (BCRA) to quantify the safe and optimal limits of repeatable behaviors. We demonstrate how hormetic alignment solves the ‘paperclip maximizer’ scenario, a thought experiment where an unregulated AI tasked with making paperclips could end up converting all matter in the universe into paperclips. Our approach may be used to help create an evolving database of ‘values’ based on the hedonic calculus of repeatable behaviors with decreasing marginal utility. Hormetic alignment offers a principled solution to the value-loading problem for repeatable behaviors, augmenting current techniques by adding temporal constraints that reflect the diminishing returns of repeated actions. It further supports weak-to-strong generalization – using weaker models to supervise stronger ones – by providing a scalable value system that enables AI to learn and respect safe behavioral bounds. This paradigm opens new research avenues for developing computational value systems that govern not only single actions but the frequency and count of repeatable behaviors.
  • Thumbnail Image
    Item
    Deep learning for low-resource machine translation : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Computer Science, School of Mathematical and Computational Sciences, Massey University, Albany, Auckland, New Zealand. EMBARGOED until further notice.
    (Massey University, 2025-09-01) Gao, Yuan
    Machine translation, a key task in natural language processing, aims to automatically translate text from one language to another while preserving semantic integrity. This thesis builds upon existing research and introduces three deep-learning methods to enhance translation performance under low-resource conditions: (i) an effective transfer learning framework that leverages knowledge from high-resource language pairs, (ii) a pre-ordering-aware training method that explicitly utilizes contextualized representations of pre-ordered sentences, and (iii) a data augmentation strategy that expands the training data size. Firstly, we develop a two-step fine-tuning (TSFT) transfer learning framework for low-resource machine translation. Due to the inherent linguistic divergence between languages in parent (high-resource language pairs) and child (low-resource language pairs) translation tasks, the parent model often serves as a suboptimal initialization point for directly fine-tuning the child model. Our TSFT framework addresses this limitation by incorporating a pre-fine-tuning stage that adapts the parent model to the child source language characteristics, improving child model initialization and overall translation quality. Secondly, we propose a training method that enables the model to learn pre-ordering knowledge and encode the word reordering information within the contextualized representation of source sentences. Pre-ordering refers to rearranging source-side words to better align with the target-side word order before translation, which helps mitigate word-order differences between languages. Existing methods typically integrate the information of pre-ordered source sentences at the token level, where each token is assigned a local representation that fails to capture broader contextual dependencies. Moreover, these methods still require pre-ordered sentences during inference, which incur additional inference costs. In contrast, our method enables the model to encode the pre-ordering information in the contextualized representations of source sentences. In addition, our method eliminates the need for pre-ordering sentences at inference time while preserving its benefits in improving translation quality. Thirdly, to address data scarcity in low-resource scenarios, we propose a data augmentation strategy that employs high-quality translation models trained bidirectionally on high-resource language pairs. This strategy generates diverse, high-fidelity pseudo-training data through systematic sentence rephrasing, generating multiple target translations for each source sentence.. The increased diversity on the target side enhances the model's robustness, as demonstrated by significant performance improvements in eight pairs of low-resource languages. Finally, we conduct an empirical study to explore the potential of applying ChatGPT for machine translation. We design a set of translation prompts incorporating various auxiliary information to assist ChatGPT in generating translations. Our findings indicate that, with carefully designed prompts, ChatGPT can achieve results comparable to those of commercial translation systems for high-resource languages. Moreover, this study establishes a foundation for future research, offering insights into prompt engineering strategies for leveraging large language models in machine translation tasks.
  • Thumbnail Image
    Item
    The challenging meet between human and artificial knowledge. A systems-based view of its influences on firms-customers interaction
    (Emerald Publishing Limited, 2023-12-18) Saviano M; Del Prete M; Mueller J; Caputo F
    Purpose This paper aims to recall the attention on a key challenge for customer relationship management related to the role of human agents in the management of the “switch point” for ensuring the effectiveness and efficiency in a customer-machine conversation. Design/methodology/approach This study contributes to the discussion about the firms’ approach to artificial intelligence (AI) in frontline interactions under the conceptual umbrella provided by knowledge management studies. Findings This paper provides a theoretical model for clarifying the role of human intelligence (HI) in AI-based frontline interactions by highlighting the relevance of the actors’ subjectivity in the dynamics and perceptions of customer-machine conversations. Originality/value An AI-HI complementarity matrix is proposed in spite of the still dominant replacement view.
  • Thumbnail Image
    Item
    Developing unbiased artificial intelligence in recruitment and selection : a processual framework : a dissertation presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Management at Massey University, Albany, Auckland, New Zealand
    (Massey University, 2022) Soleimani, Melika
    For several generations, scientists have attempted to build enhanced intelligence into computer systems. Recently, progress in developing and implementing Artificial Intelligence (AI) has quickened. AI is now attracting the attention of business and government leaders as a potential way to optimise decisions and performance across all management levels from operational to strategic. One of the business areas where AI is being used widely is the Recruitment and Selection (R&S) process. However, in spite of this tremendous growth in interest in AI, there is a serious lack of understanding of the potential impact of AI on human life, society and culture. One of the most significant issues is the danger of biases being built into the gathering and analysis of data and subsequent decision-making. Cognitive biases occur in algorithmic models by reflecting the implicit values of the humans involved in defining, coding, collecting, selecting or using data to train the algorithm. The biases can then be self-reinforcing using machine learning, causing AI to engage in ‘biased’ decisions. In order to use AI systems to guide managers in making effective decisions, unbiased AI is required. This study adopted an exploratory and qualitative research design to explore potential biases in the R&S process and how cognitive biases can be mitigated in the development of AI-Recruitment Systems (AIRS). The classic grounded theory was used to guide the study design, data gathering and analysis. Thirty-nine HR managers and AI developers globally were interviewed. The findings empirically represent the development process of AIRS, as well as technical and non-technical techniques in each stage of the process to mitigate cognitive biases. The study contributes to the theory of information system design by explaining the phase of retraining that correlates with continuous mutability in developing AI. AI is developed through retraining the machine learning models as part of the development process, which shows the mutability of the system. The learning process over many training cycles improves the algorithms’ accuracy. This study also extends the knowledge sharing concepts by highlighting the importance of HR managers’ and AI developers’ cross-functional knowledge sharing to mitigate cognitive biases in developing AIRS. Knowledge sharing in developing AIRS can occur in understanding the essential criteria for each job position, preparing datasets for training ML models, testing ML models, and giving feedback, retraining, and improving ML models. Finally, this study contributes to our understanding of the concept of AI transparency by identifying two known cognitive biases  similar-to-me bias and stereotype bias  in the R&S process that assist in assessing the ML model outcome. In addition, the AIRS process model provides a good understanding of data collection, data preparation and training and retraining the ML model and indicates the role of HR managers and AI developers to mitigate biases and their accountability for AIRS decisions. The development process of unbiased AIRS offers significant implications for the human resource field as well as other fields/industries where AI is used today, such as the education system and insurance services, to mitigate cognitive biases in the development process of AI. In addition, this study provides information about the limitations of AI systems and educates human decision makers (i.e. HR managers) to avoid building biases into their systems in the first place.
  • Thumbnail Image
    Item
    The investigation of non-contact vital signs detection microwave theoretical models and smart sensing systems : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Department of Mechanical and Electrical Engineering, SF&AT at Massey University, Palmerston North, New Zealand
    (Massey University, 2020) Nguyen, Thi Phuoc Van
    Natural disasters, such as floods, landslides and earthquakes, occur frequently around the world. The consequences of such disasters in developing countries tend to be more severe due to the lack of effective life detector systems. Life signs detecting has been an active and challenging research field that has great potential in the applications such as finding human lives under debris and non-invasive diagnosis and health monitoring. There are obvious limitations of conventional devices such as optical or acoustic detectors. The optical equipment requires operation from experts, while the acoustics need a quiet environment. The detectors with the thermal sensors and wireless tracking systems are also insufficient when the "non-line of sight" problem appears. In addition, vital signs information (such as heartbeat and breathing rate) from non-invasive microwave sensors are very important to locate people or predict health conditions in the cases of defense, smart home applications, and baby monitoring. Since NASA proposed the use of microwave radar sensing system for life detecting, research and implementation on sensitive, effective, and economic vital signs sensing systems based on microwave signals have become very active. Until now, most research on life detectors has concentrated on hardware development, signal processing, and development of new algorithms to improve accuracy of vital signs detection. The present study has focused on microwave sensors, studying microwave theoretical models and searching for life detecting, health care and smart home applications. In this research, the antennae systems for vital signs detection, such as breathing rate, were first investigated to validate their performance in a system at different frequencies. The antennae system had an extremely large band width, operating from L band to the X band. Based on the proposed antennae system, models to evaluate the false alarm/detection probabilities of a microwave sensing system were then developed and validated to examine the accuracy of the system in advance. These models are very useful for hardware development of microwave radar sensors. Further investigation into the theoretical models, proposed a novel system that was inspired by the micro bat animal's physical structure. This system showed an enhancement in the accuracy and directional signals of the microwave sensing system. Artificial intelligence was then integrated with the radar sensing system to develop the smart microwave radar sensing system. The machine learning/ deep learning models based on the collected data were developed. The study indicated high accuracy in classifying different types of breathing disorders.