---
title: "Do It, Show Me, or Help Me Decide"
canonical: "https://current.tinyfish.ai/issue/55/vision/article/65095/do-it-show-me-or-help-me-decide"
publication: "TinyFish"
publisher: "TinyFish"
language: "en-US"
guide: "https://current.tinyfish.ai/llms.txt"
author: "Sevda Polat"
author_url: "https://current.tinyfish.ai/team/200"
issue: 55
section: "vision"
section_name: "Vision"
published: 2026-11-13
word_count: 306
source_links: 1
---

# Do It, Show Me, or Help Me Decide

*By [Sevda Polat](<https://current.tinyfish.ai/team/200/markdown>)*

*November 13, 2026*

[Read on the website](<https://current.tinyfish.ai/issue/55/vision/article/65095/do-it-show-me-or-help-me-decide>)

At EMNLP, a language-technology conference, researchers last month described [eight blind people](https://aclanthology.org/2026.emnlp-main.1763.pdf) who spent three weeks working with an AI agent that could operate twelve desktop apps for them. Sometimes they asked it to finish a job. They also asked where a menu was and what an unfamiliar screen meant, so they could manage alone next time.

These are three different requests reaching one assistant. "Do it" wants a result. "Show me" wants steps the person can repeat later. "Help me decide" wants the agent to stop where the choice belongs to the person. In earlier research, an agent told to buy the cheapest sparkling water found several brands tied on price. It picked one without mentioning that their flavors differed.

Most agent benchmarks score tasks finished. That scoring counts a lesson or a pause as a miss, so every request drifts toward "do it." As these agents become everyday tools, the skill worth measuring may be hearing which request just arrived.

**Do it**
- *Cue:* "send," "fill in," "book."
- *Done well:* finishes, then says what changed so the person can check.

**Show me**
- *Cue:* "where is," "how do I."
- *Done well:* numbered keyboard steps. The screen reader, the software that speaks the screen aloud, stays in the person's app.
- *Evidence:* At CHI 2026, 12 blind users faced 2 unfamiliar Word or Excel tasks. With this help they finished 1.5 on average. With their usual tools they finished 0.5.

**Help me decide**
- *Cue:* "cheapest," "best," or a tie between options.
- *Done well:* pauses and lays out how the options actually differ.
- *Evidence:* In a 2025 test, agents that paused made choices closer to what users wanted. The tasks took longer.

**Can't tell which?**
- Ask once, before acting.
- Let each user set how often the agent checks in.

## More

- [Vision — Issue 55](<https://current.tinyfish.ai/issue/55/vision/markdown>)
- [Issue 55](<https://current.tinyfish.ai/issue/55/markdown>)
- [Publication guide](<https://current.tinyfish.ai/llms.txt>)