Trepo - Institutional Repository of Tampere University
Not a member yet
93710 research outputs found
Sort by
Sähköautojen latausmarkkinat: Markkinoiden tulevaisuudennäkymät Suomessa maksujärjestelmätoimittajan näkökulmasta Diplomityö
Get PDFTrustworthy LLMs for Ethically Aligned AI-based Systems: A PhD Research Plan
No full textIn response to growing concerns around trustworthiness and ethical alignment in AI systems, this PhD aims to investigate how Large Language Models (LLMs) can be leveraged to support ethically aligned AI development in software engineering. Despite advancements, integrating ethical principles into AI workflows remains challenging, particularly in real-world applications that require compliance with emerging regulations, such as the EU AI Act. We will develop a Visual Studio Code (VSCode) Generative AI (GenAI) Extension powered by a multi-agent LLM system with Retrieval-Augmented Generation (RAG) capabilities. The extension will be designed to aid developers by evaluating code compliance with ethical standards, providing actionable recommendations to embed trustworthiness from early stages of development. The GenAI Extension will be evaluated through an iterative design science approach, encompassing dataset generation, ethical benchmarking, and practitioner testing. A dataset of over 2000 ethically aligned AI systems, will be created in compliance with leading regulatory frameworks, serving as a foundation for this tool's assessments. With this work, we hope to assist developers, particularly in startups and SMEs, by providing practical resources for building ethically aligned AI within limited resources. Through this approach, we aim to bridge the gap between abstract ethical principles and actionable software development practices, making ethical AI more accessible across industry contexts.Peer reviewe
Developing AI strategy for high-tech industrial company
Get PDFA successful utilization of Artificial Intelligence (AI) technologies is closely linked to the development and implementation of an effective AI strategy. A well-defined AI strategy helps to identify and prioritize the most suitable AI use cases. This thesis is conducted as a case study for IONCOR, a high-tech industrial company in the battery manufacturing industry. It operates in a data-intensive environment that is highly promising for utilizing AI.
The objective of this thesis is to study: What factors should be considered when creating an AI strategy for a high-tech industrial company? The thesis establishes a theoretical foundation for AI strategy development with a literature review. Based on the literature review results, a seven-step AI strategy development framework was established. Main themes in the AI strategy devel-opment framework are: 1. Setting AI goals, 2. Use case identification and success metrics, 3. Data, 4. People and Skills, 5. Technology selection, 6. Risk analysis and Mitigation plan, and 7. Implementation.
To validate and improve the created AI strategy development framework, the empirical re-search was conducted in a case company. Data were collected through semi-structured inter-views, and the results were compared with the framework that was created. Interview themes and key questions were derived from the AI strategy development framework, and the interview data were analysed using thematic qualitative data analysis. A total of eight participants were inter-viewed in the case company based on their experience in relevant areas. Interview results aligned well with the developed AI strategy development framework. The following are the main empirical findings combined with theory.
As the first step in developing an AI strategy, it is essential to establish AI strategic objectives in collaboration with a cross-functional team and align these objectives with the company’s overall strategic goals and vision. The next step is to identify use cases that are feasible to implement and have clear business value. AI use cases should be prioritized based on value, Return On Investment (ROI), criticality, and complexity. The interview result emphasizes a focus on business value, business-critical topics, and solving business problems. The highest priority use cases should have low complexity and high value. The interview also suggested exploring whether AI presents new business opportunities. Thirdly, you need to identify the available data sources and ensure that the data is of high quality. Data must be clean, well-integrated, and easily accessible. Data security was emphasized in the interview results, along with the importance of privacy, legal, and ethical requirements. The fourth step is to evaluate the organization's current capabilities, maturity, and needed roles. Enhance the organization's AI capabilities through training and work-shops. The fifth step is to do a technology selection with the help of AI experts. The main principle in selecting technology is to select off-the-shelf solutions whenever possible, as there was a clear consensus among interviewees. AI solutions should be flexible, scalable, secure, interoperable with existing systems, and upgradable to accommodate future tasks. The sixth step is to do a risk analysis. Consider the technical, security, operational and financial risks, as well as reputational, ethical and regulatory risks. The latter ones are easy to forget based on interview feedback. The final step is to create an implementation roadmap and plan how to monitor progress. An agile and iterative approach in development should always be preferred when possible.
In conclusion, this study successfully addressed the research question by providing a practical framework for AI strategy development that can be applied in the high-tech manufacturing industry
Neurospicy Larpers: Role-play as a Method for Working with Neurodivergent Youth with Challenging Behavior Regarding Sexuality
No full textNon peer reviewe
The association between intrinsic breast cancer subtypes, mammography screening and prognosis: a large population-based real world cohort study
No full textIntroduction: Breast cancer (BC) as a heterogeneous disease is routinely managed according to its intrinsic subtypes. Mammographic BC screening reduces overall mortality in females. Our aim was to analyze the association between different intrinsic subtypes and mammography screening coverage (pre-screening-aged, screening-aged vs. post-screening-aged), attendance (attendance vs. non-attendance), and means of detection (screen-detected vs. interval BC) and BC survival. Materials and methods: We used a subpopulation of a registry including all patients diagnosed with invasive BC in Finland between 1995 and 2013. We collected screening results, information on biological characteristics and survival from national registries. Results: We included 7389 patients with early-stage BC. Compared to luminal A-like subtype, patients with triple-negative BC had the highest risks of death (HR: 1.81, 95 % CI: 1.52–2.15) and BC-related death (HR: 3.16, 95 % CI: 2.43–4.10). The majority of triple-negative BCs were diagnosed after the screening age. HER2-positive (non-luminal) tumors were most likely interval tumors, while the rest of the subtypes were most likely screen-detected. The risk of death was higher in patients with interval cancers compared to screen-detected cases (HR 1.40, 95 % CI: 1.18–1.68) and even higher among patients not attending screening (HR 2.17, 95 % CI: 1.75–2.68); this association was also detected in major subtypes. Conclusion: In this real-world dataset, triple-negative tumors had the highest risk of death and majority of these tumors were found after the screening age. In screening-aged females, patients with screen-detected tumors had the best survival, while patients with interval tumors and patients not attending screening had the worst prognosis.Peer reviewe
Distributed Home Automation with Home Assistant
No full textIn this paper, we examine implementing distributed home automation platforms, with Home Assistant serving as a case study. Home Assistant is an open-source platform that enables users to control and automate various aspects of their homes. The paper addresses the platform’s limitations due to hardware constraints, which can cause performance bottlenecks when many tasks are running at once, and with a focus on especially intensive tasks. Implementing a distributed architecture facilitates computational loads across multiple devices which improves Home Assistant’s capabilities. Key aspects include establishing a distributed network, configuring communication protocols, and developing data handling and processing methods. This approach involves working with both existing and custom-developed components and tools. The achieved results affirm the potential of a more distributed system in enhancing the performance of Home Assistant, offering a more scalable and capable solution for managing and automating various home functions.Peer reviewe
Enemy images, group experiences and propaganda in the Finnish Civil War, 1918
No full textThe Finnish Civil War of 1918 was deeply influenced by embodied rhetoric and propaganda, reflecting pre-war societal divisions between socialist aspirations and bourgeois nationalism. Reciprocal vilification intensified through propaganda. The Whites depicted themselves as guardians of independence and defenders of innocence against the formless, infected mass of the Reds, while the Reds painted the bourgeoisie as greedy oppressors and depicted the war as a leveling campaign against oppression and hunger, invoking Marxist rhetoric of class struggle. This construction of enemy images significantly lowered the threshold for violence, ultimately leading to armed conflict. Funerals became spectacles of propaganda, with the Reds emphasizing workers’ marches and the Whites invoking Christian faith and national vitality. Both sides utilized emotive language and sacrificial imagery to justify violence, with writers and journalists emerging as key propagandists. The Finnish Civil War was not just a clash of arms but a battle of narratives.Peer reviewe
Multidimensional Association Rules for Market Basket Analysis : Understanding Consumer Purchasing Patterns
Get PDFThe objective of this paper is to research whether expectation maximization clustering method and multidimensional association rules can help to identify customer purchasing patterns in high- dimensional data. The main goal is to create customer segments based on the identified patterns. The purpose is that retailers could use these segments in their operations, such as marketing, product placement in stores and product offerings.
The retail sector has been in a state of transition for a long time. Instead of the traditional, personal customer relationship, the market has been dominated by large chains. A phenomenon that has been going on in the retail sector for a long time is that small stores are becoming big grocery chains. To attract consumers, retailers have had to adapt to today's data-driven thinking to keep up with the competition.
Retailers increasingly rely on data mining techniques to identify patterns in consumer behav- ior, understand decision-making processes and plan their sales strategies accordingly. One of the most valuable tools for this purpose is Market Basket Analysis, a technique that allows retail- ers to uncover hidden patterns and relationships between products that tend to be purchased together. While many studies have examined the effectiveness of these techniques this study tries to explore customer segmentation through Expectation maximization clustering method.
The market basket data in the study contains details of the purchased items, as well as the time of purchase, making this study a multidimensional market basket analysis.
At the beginning of this paper, association analysis and clustering methods will be introduced. These form the literature review of the study. This is followed by a description of the data used in the study, the methods and finally the results.
The key findings of the study were that it was possible to identify purchasing patterns with EM clustering method and form 10 segments. The main differences between the clusters were the time of purchase and the differences between the products purchased
Amerikansuomalaisten Neuvosto-Karjalaan siirtolaiseksi pyrkivien haasteet Karjalan teknillisen avun kirjeenvaihdossa 1931-1935
Get PDFKäyttäjien sensitiivinen data femtech-sovelluksissa : Psykologisen omistajuuden näkökulma kuluttajien yksityisyyskokemukseen
Get PDFKuukautisten, hedelmällisyyden ja raskauden seurantaan suunnitellut femtech-sovellukset tarjoavat monenlaisia hyötyjä niiden käyttäjille: ne auttavat käyttäjiä seuraamaan vaivattomasti omaa vointia sekä ymmärtämään oman kehon toimintoja sekä oireita eri kierron vaiheissa, minkä lisäksi sovellukset voivat tukea käyttäjiä esimerkiksi raskauden suunnittelussa. Siten femtech-sovelluksien voidaan katsoa edistävän naisten voimaantumista. Sovellusten tarjoamien hyötyjen ohella niiden käyttöön liittyy kuitenkin tunnistettuja yksityisyysongelmia, kuten käyttäjien sensitiivisten ja henkilökohtaisten tietojen luvatonta hyödyntämistä. Näin ollen voimaantumisen ja yksityisyysongelmien välinen ristiriita luo tutkimuksellisesti mielenkiintoisen jännitteen kuluttajien sensitiivisen tiedon kokemuksen tutkimiselle kontekstissa, jossa tietojen sensitiivisyys ja henkilökohtaisuus korostuvat erityisellä tavalla.
Tässä tutkimuksessa kuluttajien sensitiivisen tiedon kokemusta valitussa kontekstissa tarkasteltiin psykologisen omistajuuden näkökulmasta, joka toimii myös osana tutkimuksen teoreettista viitekehystä. Psykologisen omistajuuden lisäksi tutkimuksen teoriaosuus koostuu yksityisyyttä käsittelevästä tutkimuksesta, jossa yksityisyyden, yksityisyyshuolien, yksityisyyskalkyylin, yksityisyysparadoksin sekä personointi-yksityisyys-paradoksin käsitteitä käsitellään tarkemmin. Siten teoreettinen viitekehys muodostuu näiden aiemman yksityisyystutkimuksen tunnistamien käsitteiden luomalle perustalle sekä teorialle psykologisesta omistajuudesta ja siihen tiiviisti kytkeytyvästä hallinnan tarpeesta, mitkä kohdistuvat käyttäjien femtech-sovellukseen syöttämiin tietoihin.
Tutkimus toteutettiin kvalitatiivisena puolistrukturoituna haastattelututkimuksena, joka koostui kahdesta eri osasta: Haastattelun ensimmäisessä osassa tutkittiin haastateltavien yleisiä ajatuksia ja kokemuksia yksityisyydestä valitussa kontekstissa, minkä lisäksi haastattelun toisessa osassa ymmärrystä yksityisyyteen liittyvistä tunteista ja käsityksistä syvennettiin hyödyntämällä narratiivista tarinantäydennysmenetelmää. Tässä tutkimuksessa tarinantäydennysmenetelmän perustana toimivat kahdeksan eri tarinaa, jotka käsittelivät monipuolisesti femtech-sovelluksen käyttöön liittyviä tilanteita. Tutkimusta varten haastateltiin seitsemää 23–29-vuotiasta naishenkilöä. Haastatteluaineisto litterointiin, minkä jälkeen aineisto analysoitiin temaattista analyysiä ja ATLAS.ti-ohjelmaa hyödyntäen.
Ensinnäkin haastatteluaineisto osoitti, että tutkimuksen kontekstissa psykologinen omistajuus ilmeni haastateltavien käyttämässä kielessä sekä tunnereaktioissa, mikä kohdistui femtech-sovellukseen syötettyihin tietoihin. Toiseksi tutkimuksen tulokset kuitenkin osoittivat, että tietojen sensitiivisyys, yksityisyys ja yksilöitävyys liittyivät siihen, kuinka voimakkaita omistajuuden tunteita haastateltavat tunsivat tietoja kohtaan. Kolmanneksi psykologiseen omistajuuteen kytköksissä olevan hallinnan tunteen voitiin todeta horjuneen tarinoiden perusteella joko tietojen luvattoman hyödyntämisen, yksityisyyden rajojen ylittämisen tai tietojen jakamiseen liittyvien riskien seurauksena, mikä herätti haastateltavissa poikkeuksetta kielteisiä tunteita. Toisaalta aineiston perusteella voitiin havaita, että oikeudenmukainen tietojen käsittely sekä luottamuksen tunne saattoivat muuttaa haastateltavien suhtautumista ja reaktioita tarinoiden tilanteisiin. Lopuksi tutkimuksen tulokset myös vahvistivat aikaisemman yksityisyystutkimuksen löydöksiä niin yksityisyyshuolien, yksityisyyskalkyylin kuin yksityisyysparadoksinkin osalta