Artificial Intelligence Exploration

I entered this week with a risk assessment mindset instead of my usual AI enthusiast perspective. I focused on selecting academic and library tasks which AI could perform responsibly to minimize time usage while avoiding situations where AI benefits did not match its potential risks. I tested these tools through low-risk operations by running them with standard test data.

Where AI Helped

• The tools for Triage & Skimming:  SciSpace and Unriddle and Answer this.  This helped me determine the value of each source in terms of time investment. step. The tool shows good performance in finding appropriate candidates and identifying essential text elements yet it does not eliminate the requirement for full human assessment. 

• Structure & De-cluttering:  The AI system Thesis and Julius provides section organization through Paperpal and Thesify which enhance written text clarity and simplicity from question sets. These tools operate as formatting tools which do not substitute for human mental operations. 

• Writer’s Block:  The scaffolds Jenni provides students help them avoid blank-page anxiety but need significant changes to match the perspective of an expert in the subject. 

Red Flags I Saw

• Source Reliability & Hallucinations:  The tools create summary versions of papers without any analytical content while producing references that appear in non-existent parts of the original documents. Any “citation” must be checked at the publisher or DOI. 

• Bias & Over-simplification:  The process of summarizing information results in two main problems which include biased interpretations and oversimplified representations of complex details. The implementation of paraphrasing tools in policy and ethics and DEI environments creates a high risk because they tend to strip away vital words and modify the core meaning of the text. 

• Privacy & Compliance:  The uploaded content together with user prompts functions as training material which could potentially be stored outside of the country. The requirements of FERPA/COPPA create a conflict with the K–12 educational environment. I believe that no PII information remains secure even when using the “private mode” feature. 

• Vendor Lock-In & Cost Drift:  The free tiers of these services create dependency while essential features become available only through paid subscriptions which results in Vendor Lock-In & Cost Drift. The operational systems of schools and libraries risk becoming unsustainable because they use tools which they cannot afford to maintain. 

• Intellectual Property & Attribution:  The process of creating Drafting tools creates confusion about who should receive credit for original work. Users can create deceptive authorship records because there are no clear notes present. 

• Equity & Access: If an assignment assumes a specific tool or paid tier, we widen the digital divide. Every AI-assisted path needs a no-AI alternative. 

Tool-by-Job Takeaways

• The initial list generation process of Literature discovery (Consensus, Elicit) works well but it does not produce systematic search results effectively. Expect gaps; verify everything. 
• The PDF Q&A/summarization tools from SciSpace and Unriddle and AnswerThis provide a good starting point but they might fail to detect subtle details and they could interpret tables and figures incorrectly. Treat as directional. 
• The process of drafting and outlining (Thesis AI, Julius, Jenni) helps with organizing the structure of the work but does not enhance the actual content. Risk of tone drift and invented claims—revise heavily. 
• The editing/paraphrasing tools (Paperpal and Thesify) assist with clarity but they may eliminate the precise language of technical terms. The following text requires precise language to be reintroduced after the pass. 

Final Position

AI functions as a power stapler in specific situations by providing fast assembly of components. The tool operates as a reference system yet it cannot substitute the necessity for expert knowledge and reading abilities and ethical choices. The most secure approach for educational institutions and libraries requires a skeptical approach to adoption which involves limited specific applications with complete verification processes and complete disclosure of information.